contentsmat1.uibk.ac.at/download/oemg2019/oemg2019.pdf · contents welcome 9 sponsor and...
TRANSCRIPT
Contents
Welcome 9
Sponsor and Cooperation Partners 11
Conference Opening 13
General Information 15Conference Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Conference and Registration Desk . . . . . . . . . . . . . . . . . . . . . 15Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Technical Equipment of the Lecture Rooms . . . . . . . . . . . . . . . . 16Internet Access during the Conference . . . . . . . . . . . . . . . . . . . 16Lunch and Dinner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Coffee Breaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Getting there and around . . . . . . . . . . . . . . . . . . . . . . . . . . 17Information about the Conference Venue Dornbirn . . . . . . . . . . . . . 19Information about the University of Applied Sciences Vorarlberg . . . . . 19Conference Venue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Conference Organization and Committees 21Program Committee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Local Organizing Committee . . . . . . . . . . . . . . . . . . . . . . . . 21Sections Coordinators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Organizers of Minisymposia . . . . . . . . . . . . . . . . . . . . . . . . 22LehrerInnen-Tag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Hochschul-Tag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Meetings and Public Program 25General Assembly, ÖMG . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1
2 Contents
Public Lecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Side Events 27LehrerInnentag 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Hochschultag 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Social Program 29Reception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Conference Dinner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Public Lecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Excursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Detailed Program of Sections and Minisymposia 31Timetable Monday, September 16, Afternoon Sessions . . . . . . . . . . 32Timetable Monday, September 16, Evening Sessions . . . . . . . . . . . 33Timetable Tuesday, September 17, Morning Sessions . . . . . . . . . . . 34Timetable Tuesday, September 17, Afternoon Sessions . . . . . . . . . . 35Timetable Tuesday, September 17, Evening Sessions . . . . . . . . . . . 36Timetable Wednesday, September 18, Morning Sessions . . . . . . . . . 37Timetable Thursday, September 19, Morning Sessions . . . . . . . . . . . 38Timetable Thursday, September 19, Afternoon Sessions . . . . . . . . . . 39Timetable Thursday, September 19, Evening Sessions . . . . . . . . . . . 40
ABSTRACTS 41
Plenary Speakers 43Efstathia Bura: Sufficient Dimension Reduction in high dimensional re-
gression and classification . . . . . . . . . . . . . . . . . . . . . . 44Christopher Frei: Average bounds for the `-torsion in class groups of num-
ber fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Fritz Gesztesy: The Birman sequence of integral inequalities and some of
its refinements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Marlis Hochbruck: On leapfrog-Chebyshev schemes . . . . . . . . . . . . 45Josef Hofbauer: Permanence of biological systems . . . . . . . . . . . . . 45Ozlem Imamoglu: Binary quadratic forms, quadratic fields and Modular
forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Helmut Pottmann: Discrete geometric structures motivated by architecture
and computational design . . . . . . . . . . . . . . . . . . . . . . . 46Wilhelm Schlag: On no-return theorems in nonlinear dispersive PDEs . . 47Susanna Terracini: Spiralling and other solutions in limiting profiles of
competition-diffusion systems . . . . . . . . . . . . . . . . . . . . . 48
Contents 3
M1: Industrial and Applied Mathematics 49Christoph Erath: Efficient solving of a time-dependent interface problem . 50Klaus Frick: Using extended Kalman Filters for solving non-linear inverse
problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51Richard Kowar: Iterative methods for PAT in dissipative media . . . . . . 51Gerhard Zangerl: Full field inversion in photoacoustic tomography with
variable sound speed . . . . . . . . . . . . . . . . . . . . . . . . . 52Barbara Kaltenbacher: Parameter identification and uncertainty quantifi-
cation in stochastic state space models and its application to textureanalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Lukas Neumann: Block coordinate descent method for ill-posed problems 53Michael Liebrecht: Process Model for the Industrial Plate Leveling Oper-
ation with Special Emphasis on Online Control . . . . . . . . . . . 54Dimitri Rothermel: Nonlinear inverse heat transfer problems in modelling
the cooling process of heavy plates . . . . . . . . . . . . . . . . . . 54Philip Sellars: Hyperspectral Image classification with Minimal Learning:
A Graph Based Approach . . . . . . . . . . . . . . . . . . . . . . . 55Johannes Schwab: Regularizing Networks for Solving Inverse Problems . 55Andreas Maier: Known Operator Learning U An Approach to unite
Physics, Signal Processing, and Machine Learning . . . . . . . . . 56
M2: Mathematical Finance in the age of Machine Learning 57Juan-Pablo Ortega: Dynamic and Control Theoretical Aspects of Reservoir
Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58Lyudmila Grigoryeva: Forecasting of Realized (Co)Variances with Reser-
voir Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Lukas Gonon: Error Bounds for Random Recurrent Neural Networks and
General Reservoir Computing Systems . . . . . . . . . . . . . . . . 59Josef Teichmann: Learning stochastic dynamics by random signatures
with applications to mathematical Finance . . . . . . . . . . . . . . 59Anastasis Kratsios: Universal Approximation Theorems . . . . . . . . . . 60Hanna Sophia Wutte: Randomized shallow neural networks and their use
in understanding gradient descent . . . . . . . . . . . . . . . . . . 60Christa Cuchiero: Modeling rough covariance processes . . . . . . . . . 61Wahid Khosrawi: A Neural Network Approach to Local Stochastic Volatil-
ity Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Thomas Krabichler: Deep ALM . . . . . . . . . . . . . . . . . . . . . . . 63Sara Svaluto-Ferro: Infinite dimensional polynomial jump-diffusions . . . 63
M3: Parallel and Distributed Optimization and Simulation 65
4 Contents
Hayk Shoukourian: Predicting the efficiency of LRZ’s cooling infrastruc-ture using machine learning . . . . . . . . . . . . . . . . . . . . . 66
Philipp Gschwandtner: Towards Multi-objective, Region-based Auto-tuning for Parallel Programs . . . . . . . . . . . . . . . . . . . . . 67
Thomas Feilhauer: Optimization with the Distributed Execution Framework 68Klaus Rheinberger: Analyzing Autonomous Demand Side Algorithms with
Parallel Computation Frameworks . . . . . . . . . . . . . . . . . . 69
M4: Spectral Theory 71Thomas Kappeler: On the integrability of the Benjamin-Ono equation with
periodic boundary conditions . . . . . . . . . . . . . . . . . . . . . 72Daniel Lenz: Uniformity of cocycles and spectra of Schreier graphs . . . 72Aleksey Kostenko: On the absolutely continuous spectrum of generalized
indefinite strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72Roland Donninger: Strichartz estimates for the one-dimensional wave
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73Alexander Sakhnovich: Discrete Dirac systems: spectral and scattering
theories and Verblunsky-type coefficients . . . . . . . . . . . . . . . 73Gueorgui Dimitrov Raykov: Threshold Singularities of the Spectral Shift
Function for Geometric Perturbations of a Magnetic Hamiltonian . 74Johanna Michor: Rarefaction and shock waves for the Toda equation . . . 74Mateusz Piorkowski: Riemann-Hilbert approach to asymptotic analysis . 75Rúben Sousa: Product formulas and convolutions for solutions of Sturm-
Liouville equations . . . . . . . . . . . . . . . . . . . . . . . . . . 75Ziping Rao: Optimal Blowup Stability for the Energy Critical Wave Equation 76
M5: Inverse Problems for Partial Differential Equations 77Markus Haltmeier: NETT: Solving Inverse Problems with Deep Neural
Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78Daniel Obmann: Sparse synthesis regularization using deep neural networks 78Ronny Ramlau: Singular Value Decomposition for Atmospheric Tomography 78William Rundell: Determining nonlinear terms in a reaction-diffusion system 79Frank Hettlich: Identification and Design of EM-Chiral Scattering Obstacles 80Alexey Agaltsov: Uniqueness and reconstruction in a passive inverse
problem of helioseismology . . . . . . . . . . . . . . . . . . . . . . 80Anne Wald: Parameter identification for the Landau-Lifshitz-Gilbert
equation in Magnetic Particle Imaging . . . . . . . . . . . . . . . . 81Gwenael Mercier: On convergence of Total Variation regularized inverse
problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Iwona Piotrowska-Kurczewski: Generalized tolerance regularization . . . 83
Contents 5
Francisco Romero Hinrichsen: Dynamical super-resolution with applica-tions to Ultrafast ultrasound imaging . . . . . . . . . . . . . . . . . 84
Rebecca Klein: Regularizing sequential subspace optimization for theidentification of the stored energy of a hyperelastic structure . . . . 84
Sarah Eberle: Solving the inverse problem of linear elasticity with mono-tonicity methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
M6: Enumerative and Algebraic Combinatorics 87Wenjie Fang: The Steep-Bounce Zeta Map in Parabolic Cataland . . . . . 88Alexander R. Miller: Some results on characters of symmetric groups . . 88Michael Joseph Schlosser: A weight-dependent inversion statistic and
Catalan numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 89Manuel Kauers: Making Many More Matrix Multiplication Methods . . . 90Benedikt Stufler: Local convergence of random planar graphs . . . . . . 90Emma Yu Jin: Exact and asymptotic enumeration of ascent sequences and
Fishburn matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
S01: Logic, Set Theory, and Theoretical Computer Science 93Marlene Koelbing: Distributivity of forcing notions . . . . . . . . . . . . 94Johannes Philipp Schürz: Strong measure zero sets in the higher Cantor
space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94Jonathan Schilhan: The tower spectrum . . . . . . . . . . . . . . . . . . 95Lorenz Halbeisen: Formen des Auswahlaxioms in der Ringtheorie . . . . 95Victor Torres-Perez: Diamonds, games and cardinal invariants . . . . . . 95Moritz Müller: Automating Resolution is NP-Hard . . . . . . . . . . . . 96Salome Schumacher: The relation between two weak choice principles . . 96Vera Fischer: Independence and almost disjointness . . . . . . . . . . . . 97Wolfgang Wohofsky: Cardinal characteristics and the Borel Conjecture . 98
S02: Algebra, Discrete Mathematics 99Alois Panholzer: Label-quantities in trees and mappings . . . . . . . . . 100Daniel Krenn: Optimal Multi-Pivot Quicksort and its Asymptotic Analysis 100Michael Wibmer: Galois groups of differential equations . . . . . . . . . 100
S03: Number Theory 101Kevin Destagnol: Counting points of given degree and bounded height via
the height zeta function . . . . . . . . . . . . . . . . . . . . . . . . 102Nick Rome: On Serre’s Problem for Conics . . . . . . . . . . . . . . . . 102Christopher Frei: Arithmetic progressions in binary quadratic forms and
norm forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102Shuntaro Yamagishi: Solving equations in many variables in primes . . . 103
6 Contents
S04: Algebraic Geometry, Convex and Discrete Geometry 105Hassan Mahmoud Al-Zoubi: ANCHOR RINGS OF FINITE TYPE GAUSS
MAP IN THE EUCLIDEAN 3-SPACE . . . . . . . . . . . . . . . . 106
S05: Computational Geometry and Topology 107Roland Kwitt: Deep Homological Learning . . . . . . . . . . . . . . . . 108Daniela Egas Santander: Topological explorations in neuroscience . . . . 108Alexander Scott Rolle: Spaces of clusterings . . . . . . . . . . . . . . . . 109Anton Nikitenko: Stochastic applications of discrete topology . . . . . . 109Ziga Virk: Detecting lengths of geodesics with higher-dimensional persis-
tence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Jie Liang: Topological Structures of Probabilistic Landscape of Stochastic
Reaction Networks . . . . . . . . . . . . . . . . . . . . . . . . . . 110Ligia Loretta Cristea: On fractal geometric and topologic properties of
triangular labyrinth fractals . . . . . . . . . . . . . . . . . . . . . 111Gunther Leobacher: Necessary and sufficient conditions for the local
unique nearest point property of a real sub-manifold . . . . . . . . 111Adam Brown: Convergence and Stability of Random Mapper . . . . . . . 112Ulrich Bauer: Dualities and clearing for image persistence . . . . . . . . 112
S06: Ordinary Differential Equations and Dynamical Systems 113Peter Szmolyan: Periodic solutions and and switching behavior in a model
of bipolar disorder . . . . . . . . . . . . . . . . . . . . . . . . . . 114Yacine Halim: On a system of difference equations of second order solved
in a closed from . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Chems Eddine Arroud: Optimality conditions of Moreau sweeping process 116
S07: Partial Differential Equations and Calculus of Variations 117Birgit Schörkhuber: Nonlinear asymptotic stability of homothetically
shrinking Yang-Mills solitons . . . . . . . . . . . . . . . . . . . . . 118Reinhard Bürger: Two-locus clines maintained by diffusion and recombi-
nation in a heterogeneous environment . . . . . . . . . . . . . . . . 118Simon Blatt: The negative L2 gradient flow for p-elastic energies . . . . . 119Anna Kiesenhofer: Small data global regularity for half-wave maps in 4
dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119Irfan Glogic: Threshold for blowup for the supercritical cubic wave equation120Nicole Vorderobermeier: On the analyticity of critical points of the Möbius
energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120Lucrezia Cossetti: Unique Continuation for the Zakharov Kuznetsov equa-
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Contents 7
Luigi Forcella: Interpolation theory and regularity for incompressiblefluid models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
David Sebastian Wallauch: Stable blowup for a nonlinear Klein-Gordonequation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Ani Tumanyan: Fredholm property of regular hypoelliptic operators inmultianisotropic Sobolev spaces . . . . . . . . . . . . . . . . . . . 122
S08: Functional Analysis, Real and Complex Analysis 123Martin Kolar: Infinitesimal symmetries of weakly pseudoconvex manifolds 124Michael Reiter: The reflection map of sphere mappings . . . . . . . . . . 124Gerhard Racher: On Fourier multipliers . . . . . . . . . . . . . . . . . . 124Christian Bargetz: On Generic Properties of Nonexpansive Mappings . . 125Eduard Nigsch: The geometrization of the theory of full Colombeau algebras125Michael Dymond: Highly irregular separated nets. . . . . . . . . . . . . 125Ilya Kossovskiy: Classification of homogeneous strictly pseudoconvex hy-
persurfaces in C3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 126Friedrich Haslinger: The d-complex on the Segal-Bargmann space . . . . 126Karlheinz Groechenig: Sampling, Marcinkiewicz-Zygmund sets, approxi-
mation, and quadrature rules . . . . . . . . . . . . . . . . . . . . . 127Stefan Haller: The heat kernel expansion of Rockland differential opera-
tors and applications to generic rank two distributions in dimensionfive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Umi Mahnuna Hanung: A role of Harnack extension in integrating func-tions over arbitrary subsets . . . . . . . . . . . . . . . . . . . . . . 129
Uri Grupel: Intersections with random geodesics in high dimensions . . . 130
S09: Numerical Analysis, Scientific Computing 131Olaf Steinbach: Space-Time Finite Element Methods . . . . . . . . . . . 132Robert Eberle: Vibration of a beam under structural randomness . . . . . 133Teodora Catinas: On some interpolation operators on triangles with
curved sides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134Markus Melenk: an adaptive algorithm for the fractional Laplacian . . . 135Somveer Singh: A new approach of operational matrices for hyperbolic
partial differential equations . . . . . . . . . . . . . . . . . . . . . 135Anand Kumar Yadav: Effect of Impedance on Reflection of Plane Waves
in a Rotating Magneto-thermoelastic Solid Half-Space with Diffusion 136Vinita Devi: An operational matrix approach for two-dimensional hyper-
bolic telegraph equation . . . . . . . . . . . . . . . . . . . . . . . 136Rahul Kumar Maurya: Multistep finite difference scheme for electromag-
netic waves model arising from dielectric media . . . . . . . . . . . 137
8 Contents
S10: Optimization 139Issam A.R. Moghrabi: A Multi-step Hybrid Conjugate Gradient Method . 140
S11: Probability, Statistics 141Thomas Fetz: Computing upper probabilities of failure using Monte Carlo
simulation, reweighting techniques and fixed point iteration. . . . . 142Alexander Steinicke: BSDEs with Jumps in the Lp-setting: Existence,
Uniqueness and Comparison . . . . . . . . . . . . . . . . . . . . . 143Christian Pfeifer: Embedding an empirically based decision strategy for
backcountry skiers into a subjective probability framework . . . . . 144Wolfgang Woess: Recurrence of 2-dimensional queueing processes, and
random walk exit times from the quadrant . . . . . . . . . . . . . . 145Ercan Sönmez: Random Walk on the Random Connection Model . . . . . 145
S12: Financial and Actuarial Mathematics 147Stefan Rigger: Interacting Particle Systems, Default Cascades and the
M1-topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148Jörn Sass: Utility Maximizing Strategies under Increasing Model Uncer-
tainty in a Multivariate Black Scholes Type Market . . . . . . . . . 149Sascha Desmettre: Severity Modeling of Extreme Insurance Claims for
Tariffication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150Josef Anton Strini: Optimal dividends and funding in risk theory . . . . . 151Stefan Thonhauser: Optimal reinsurance for expected penalty functions in
risk theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
List of Participants 153
Index 157
Welcome
The board of the Austrian Mathematical Society (’OMG) and the local conferenceorganizers welcome you to the ÖMG-Conference in Dornbirn, September 16–20,2019. The meeting takes place in the lecture rooms of the
University of Applied Sciences Vorarlberg in Dornbirn.
The scientific program starts on Monday, September 16, at 9:00 in room W211/12and continues until the evening of Thursday, September 19, with an additional day onFriday, September 20, featuring the Teacher’s Day and the Universities of AppliedSciences and Education Day (Hochschul-Tag).
The scientific program consists of:
A plenary lecture of the ÖMG Award Winner 2019: Christopher Frei (University ofManchester).
Furthermore, 8 invited plenary lectures including Wilhelm Schlag (Yale Univer-sity), Helmut Pottmann (KAUST and TU Wien), Marlis Hochbruck (KIT Karlsruhe),Özlem Imamoglu (ETH Zürich), Efstathia Bura (TU Wien), Josef Hofbauer (Univer-sity of Wien), Susanna Terracini (University of Turin), and Fritz Gesztesy (BaylorUniversity).
The program is then split up into 12 sections and 6 minisymposia with altogether125 talks.
Moreover, a public lecture will be given by Karl Sigmund on Thursday evening,September 19, at 19:00 in room W211/12.
9
10 Welcome
The list of participants includes about 200 people coming from Algeria, Armenia,Austria, Chile, Czech Republic, France, Germany, Hungary, India, Indonesia, Italy,Jordan, Kuwait, Mexico, Netherlands, Poland, Portugal, Romania, Saudi Arabia,Slovenia, Spain, Switzerland, United Kingdom, and United States.
The conference languages are English and German.
The scientific program is complemented by several social activities. Besides thecoffee breaks there is a reception on Monday evening, a free Wednesday afternoonto enjoy the excursion to the beautiful “Bregenzer Wald”. The conference dinneris included in the participation fee and is scheduled for Wednesday evening at theKarren.
In this book you can find useful information, the program day-by-day and also or-dered for each section and minisymposia. Moreover, all abstracts are contained here.For additional information and updates visit the conference webpagehttp://www.oemg.ac.at/Tagungen/2019/We hope that you will have a very fruitful and enjoyable time in Dornbirn.
K. U. would like to thank all section and minisymposia organizers for their effort andsupport. Special thanks to Thomas Fetz for help with the program booklet, ClemensHeuberger for assistance with the website, and finally Clemens Fuchs for supportand permission to use his templates and all resources!
Dornbirn, September 2019 The Local Organizers
Sponsor and CooperationPartners
We thank our sponsors and cooperation partners who provided generous support forthe conference: the University of Applied Sciences Vorarlberg and the Governmentof Vorarlberg.
DORNBIRN!!!!!!!!!!!!!
11
12 Sponsor and Cooperation Partners
Conference Opening
The opening ceremony takes place in room W211/12 on Monday, September 16,9:00-9:30. The inaugural addresses will be given by:
• Prof. Dr. Karl Unterkofler (Chairman of the Local Organizing Committee),
• Prof. Dr. Barbara Kaltenbacher (President of the Austrian Mathematical Society).
The scientific program starts on Monday, September 16, at 9:30 with the plenarylecture of Wilhelm Schlag (Yale University) entitled “On no-return theorems in non-linear dispersive PDEs”.
After the first plenary lecture there will be a coffee break from 10:30 to 11:00.
The second plenary lecture entitled “Discrete geometric structures motivated by ar-chitecture and computational design” will be given by Helmut Pottmann (KAUSTand TU Wien). This lecture concludes the Monday morning.
13
14 Conference Opening
General Information
Conference Location
The conference takes place in the lecture rooms of the University of Applied SciencesVorarlberg (FHV). The address is:
Hochschulstraße 16850 DornbirnAustria
The FHV can be contacted by phone at +43 (0)5572 792 0.
Conference and Registration Desk
The conference desk is located in room W207/08 on the 3rd floor (2. Stock). Itsopening hours are:
Monday, September 16, 08:00–12:00 and 14:00–17:00,Tuesday, September 17, 08:15–11:00,Wednesday, September 18, 08:15–11:00,Thursday, September 19, 08:15–11:00,Friday, September 20, 08:15–11:00.
E-Mail: [email protected]
Registration
Registration of participants takes place on Monday, September 16, 2019, 08:00–12:00 and 14:00–17:00 in room W207/08 as well as during the opening hours of theconference desk from Monday–Friday.
15
16 General Information
Technical Equipment of the Lecture Rooms
All rooms are equipped with PC and beamer. The lecture rooms are equipped withlarge blackboards. Technical assistants are present to help you transfer your file tothe PC.
Internet Access during the Conference
WLAN access will be available to conference participants. Access data can be foundin the conference folder. There is also a good coverage for the eduroam networkacross the conference venue.
Lunch and Dinner
Lunch and a coffee or other refreshments in between can be taken (on one’s ownexpense) at the FHV Mensa & Cafe Schräg which is available at the venue of thecongress. Please note that the semester already started and the capacity is restricted.
Alternatively you can visit the restaurants and shops nearby:
EUROSPAR, Marktstraße 67,https://www.spar.at/standorte/eurospar-dornbirn-6850-marktstrasse-67
Gasthaus Gemsle, Marktstrasse 62,http://www.gemsle.at
Shao Kao, Hintere AchmühlerstraSSe 1,https://shaokao.at/
La Scarpetta Ristorante - Pizzeria, Hintere Achmühlerstrasse 1a,http://www.ristorante-lascarpetta.at
Restaurant Krone, Hatlerstrasse 2,https://www.kronehotel.at/hotel-dornbirn/restaurant
Bella Napoli, Schillerstrasse 1,
Gabriel’s Cucina, Marktstrasse 14,http://www.gabrielscucina.at
Bierlokal, Marktstrasse 12/2,http://www.bierlokal.at
21 Steak & Fisch, Marktstrasse 21,https://www.steakhaus21.at
General Information 17
Rotes Haus, Marktplatz 13,http://roteshaus.at
Pizzeria San Marco, Marktplatz 12,http://sanmarco-dornbirn.at
On Monday evening there will be a reception at 7pm at the foyer of the Fach-hochschule. You are invited free of charge to these event. Please be prepared thatthere will be only a small appetizer and that this will not replace your dinner.
The conference dinner takes place on Wednesday evening at 6.30pm at Karrenrestaurant, Dornbirn. It is included in the conference fee. You can either take thecable car (included too) or take part in a guided walk (elevation up 500m).
Coffee Breaks
Refreshments (coffee, juices, mineral water, cookies) are served free of charge toparticipants on:
Monday, 10:30–11:00 and 14:45–15:15,Tuesday, 09:30–10:00 and 15:00–15:30,Wednesday, 09:30–10:00,Thursday, 09:30–10:00 and 15:00–15:30,Friday, 10:15–10:45 and 14:30–15:00.
In addition and for self-payment, the university cafeteria serves coffee and snacks.
Getting there and around
(i) How to reach Dornbirn (international travel)
By plane
The nearest international airport is Zürich (Switzerland, 120 km from Dornbirn), thenearest local airport is Altenrhein (Switzerland, 20 km from Dornbirn)
(ii) How to reach the Fachhochschule Vorarlberg
The public transport network is very well developed in Dornbirn and Vorarlberg.Dornbirn can be easily reached by train and car. Dornbirn offers a comprehensive bus
18 General Information
network. Due to its central location in Dornbirn, Vorarlberg University of AppliedSciences (FH Vorarlberg) can also be reached easily on foot.
By Train
When traveling by train, we recommend that you leave the train at Dornbirn maintrain station (Hauptbahnhof). From there you can take a bus or walk to FH Vorarl-berg. The walk takes 15 to 20 minutes.
Train connections - Österreichische Bundesbahnen - Austrian Federal Railways:https://www.oebb.at/en
By public buses
From the central bus station (located in front of the main train station in Dornbirn)you can take the following buses to reach FH Vorarlberg. The trip takes about 5minutes; during the day buses run every 15 minutes.
• Line 50 (yellow buses) - Direction of “Lustenau-Gaissau” (exit at "Sägerbrücke"bus stop)
• Line 22 and 23 (yellow buses) - Direction of “Götzis” (exit at "Sägerbrücke" busstop)
• Line 2 (red buses) - Direction of “Brehmenmahd” (exit at "Sägerbrücke" bus stop)
• Line 3 (red buses) - Direction of “In Fängen” (exit at "Sägerbrücke" bus stop)
Bus connections and app for tickets: https://www.vmobil.at
By car
Take the Rheintalautobahn to the motorway exit of Dornbirn Süd and then followthe signs to the city centre (Zentrum). Turn left after the 5th set of traffic lights (atDornbirn hospital), then right after the 2nd set of traffic lights and follow the signsto the "Fachhochschule Vorarlberg."
Parking at FH Vorarlberg:https://www.fhv.at/fileadmin/user_upload/fhv/files/anreise/Parkmoeglichkeiten_an_der_FH_Vorarlberg.pdf
General Information 19
Information about the Conference Venue Dornbirn
Dornbirn is the largest town in Vorarlberg. For more information about its historyand culture, please visit the web page: https://www.dornbirn.info
Information about the University of Applied Sciences Vorarlberg
The University of Applied Sciences Vorarlberg was founded in 1992. For moreinformation, please visit the web page: https://www.fhv.at
20 General Information
Conference Venue
Hochschulstraße 1, 2. Stock:Rooms W205, W206, W211/12, U210, W207/08 (registration).
Hochschulstraße 1, 4. Stock: Rooms U405, U406, U407.
Hochschulstraße, 2. Obergeschoss
Hochschulstraße, 4. ObergeschossU4 05
U2 10
W2 05
W2 11/12
W2 07/08
W2 06
U4 06 U4 07
Conference Organization andCommittees
Program Committee
András Bátkai (Vorarlberg)Mathias Beiglboeck (Wien)Reinhard Bürger (Wien)Michael Eichmair (Wien)Barbara Kaltenbacher (Klagenfurt)Alexander Ostermann (Innsbruck)Hans-Peter Schröcker (Innsbruck)Gerald Teschl (Wien)Robert Tichy (Graz)Karl Unterkofler (Vorarlberg, Chair)
Local Organizing Committee
Thomas Fetz (Innsbruck)Steffen FinckMichael HellwigKathi PlankensteinerCarmen Platonina (Organization management)Natascha Senoner (Secretary)Karl Unterkofler (Chair)
21
22 Conference Organization and Committees
Sections Coordinators
• S01. Logic, Set Theory, and Theoretical Computer ScienceVera Fischer (Wien), Martin Goldstern (Wien)
• S02. Algebra, Discrete MathematicsClemens Heuberger (Klagenfurt), Clemens Fuchs (Salzburg)
• S03. Number TheoryTimothy Browning (IST Austria)
• S04. Algebraic Geometry, Convex and Discrete GeometryTim Netzer (Innsbruck)
• S05. Computational Geometry and TopologyHerbert Edelsbrunner (IST Austria), Michael Kerber (Graz)
• S06. Ordinary Differential Equations and Dynamical SystemsPeter Szmolyan (Wien)
• S07. Partial Differential Equations and Calculus of VariationsRoland Donninger (Wien)
• S08. Functional Analysis, Real and Complex AnalysisEva Kopecka (Innsbruck), Bernhard Lamel (Wien)
• S09. Numerical Analysis, Scientific ComputingMarkus Melenk (Wien), Olaf Steinbach (Graz)
• S10. OptimizationRadu Bot (Wien)
• S11. Probability, StatisticsEvelyn Buckwar (Linz), Michaela Szölgyenyi (Klagenfurt)
• S12. Financial and Actuarial MathematicsGunther Leobacher (Graz), Stefan Thonhauser (Graz)
Organizers of Minisymposia
• M1. Industrial and Applied Mathematics:Organizer: Markus Haltmeier (UNI Innsbruck)
• M2. Mathematical Finance in the age of Machine Learning:Organizer: Josef Teichmann (ETH Zürich)
• M3. Parallel and Distributed Optimization and Simulation:Organizer: Steffen Finck (FH Vorarlberg)
Conference Organization and Committees 23
• M4. Spectral Theory:Organizers: Gerald Teschl (UNI Wien), Jussi Behrndt (TU Graz)
• M5. Inverse Problems for Partial Differential Equations:Organizers: William Rundell (Texas A&M University), Barbara Kaltenbacher(UNI Klagenfurt)
• M6. Enumerative and Algebraic Combinatorics:Organizers: Michael Drmota (TU Wien), Christian Krattenthaler (UNI Wien)
LehrerInnen-Tag
Bernd Thaller (Universität Graz)András Batkái (Pädagogische Hochschule Vorarlberg)
Hochschul-Tag
Karl Unterkofler (Dornbirn)Susanne Teschl (Wien)
24 Conference Organization and Committees
Meetings and Public Program
General Assembly, OMG
The General Assembly of the Austrian Mathematical Society will not take placeduring the conference. Instead it is planned for November 22nd in Graz during theAustrian Mathematics Day (Tag der Mathematik 2019).
Public Lecture
Participants of the conference are invited to attend the public lecture on Thursday,September 18, at 7pm in room W211/12. The lecture is open to the general publicand will be given in German.
25
26 Meetings and Public Program
Side Events
LehrerInnentag 2019
Der LehrerInnentag 2019 findet im Rahmen der ÖMG-Konferenz 2019 am Freitagden 20. September an der FH Vorarlberg im Hörsaal W211/12 statt. Die Teilnahmeam LehrerInnentag ist kostenfrei.
Der LehrerInnentag widmet sich dem Thema Computereinsatz im Mathematikunter-richt - Top oder Flop?
http://www.oemg.ac.at/Tagungen/2019/index.php/program/lehrerinnentag-2019
Hochschultag 2019
Der Hochschultag 2019 (ehemals FH-Tag) findet am Freitag den 20. Septemberim Rahmen der ÖMG-Konferenz 2019 an der FH Vorarlberg im Hörsaal W206statt. Er ist als Fortbildungstag für Mathematiklehrende an Fach- und PädagogischenHochschulen konzipiert.
Für die Teilnahme (nur am Hochschultag) ist ein Unkostenbeitrag von Euro 60.-vorgesehen; die Anmeldung erfolgt über die Registrierungsplattform der Konferenz.Für Konferenz-Teilnehmer ist die Veranstaltung kostenfrei.
Die Themenkreise umfassen verschiedene Aspekte der Digitalisierung im Mathe-matikunterricht.http://www.oemg.ac.at/Tagungen/2019/index.php/program/hochschultag-2019
27
28 Side Events
Social Program
In the following you will find a list of social events and excursions offered through-out the week. The conference excursion on Wednesday in the afternoon is foreverybody (facultative) and can be booked through the registration platform of thecongress.
Reception
There will be a reception on Monday evening at 7pm at the Foyer of the Fach-hochschule Vorarlberg sponsored by Stadt Dornbirn and Land Vorarlberg.
A welcome address will be delivered by the Rector of the Fachhochschule VorarlbergProf. Dr. Tanja Eisenle and a representative of the regional government authorityMag. Gabriela Dür.
Conference Dinner
The conference dinner on Wednesday evening is included in the conference fee. Itwill take place at Karren restaurant, Dornbirn and starts at 6.30 PM.You can either take the cable car (included too) or take part in a guided walk (eleva-tion up 500m).
Public Lecture
A public lecture entitled Open Code - Die Mathematik als Universalsprache will begiven by Karl Sigmund (UNI Wien) on Thursday evening at 7pm.
29
30 Social Program
Zum Vortrag: Die Mathematik wird oft mit einer Sprache verglichen. Ihr Verhältniszur Umgangssprache ist vielschichtig, was zu einem ganzen Spektrum von Codierun-gen führt. Außerdem ist die Sprache der Mathematik auf zweierlei Art universell:sie wird überall gesprochen - wohl auch von extragalaktischen Intelligenzen, falls eswelche gibt - und sie ist auf alles anwendbar, worüber man exakt denken kann, sogarauf sich selbst.
Excursion
The following excursion will take place on Wednesday afternoon and can also bebooked directly by the registration procedure:
• Tour: Bregenzer Wald, Käsekeller Lingenau. Price: Euro 29.-
Further recommended self-organized tours:
• Visit Bregenzhttps://www.bregenz.travel
• Visit the Rolls Royce museum in Dornbirnhttps://www.rolls-royce-museum.at
• Hike through the Rappenlochschluchthttps://www.karren.at/wandern/rappenlochschlucht-8/
Detailed Program of Sectionsand Minisymposia
31
32 Detailed Program of Sections and Minisymposia
Mon
day,
Sep
tem
ber
16,A
fter
noon
Ses
sion
s
15:1
5
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
M3
a.1
H.S
ho
uk
ou
ria
nP
redic
ting
the
effi
cien
cyof
LR
Z’s
cooli
ng
infr
astr
uct
ure
usi
ng
mac
hin
ele
arnin
g.
S1
1a.
1T
.F
etz
Com
puti
ng
upper
pro
bab
ilit
ies
of
fail
ure
usi
ng
Monte
Car
losi
mula
tion,
rew
eighti
ng..
.
M1
a.1
C.E
ra
thE
ffici
ent
solv
ing
of
a
tim
e-dep
enden
tin
terf
ace
pro
ble
m.
S0
8a.
2M
.R
eit
er
The
refl
ecti
on
map
of
spher
e
map
pin
gs.
S0
8a.
3G
.R
ach
er
On
Fouri
erm
ult
ipli
ers.
M1
a.3
R.K
ow
ar
Iter
ativ
em
ethods
for
PA
Tin
dis
sipat
ive
med
ia.
M1
a.2
K.F
ric
kU
sing
exte
nded
Kal
man
Fil
ters
for
solv
ing
non-l
inea
r
inver
sepro
ble
ms.
S1
1a.
2A
.S
tein
ick
eB
SD
Es
wit
hJu
mps
inth
e
Lp-s
etti
ng:
Exis
tence
,
Uniq
uen
ess
and
Com
par
ison.
M3
a.3
T.F
eil
ha
uer
Opti
miz
atio
nw
ith
the
Dis
trib
ute
dE
xec
uti
on
Fra
mew
ork
.
M3
a.2
P.G
sch
wa
nd
tner
Tow
ards
Mult
i-obje
ctiv
e,
Reg
ion-b
ased
Auto
-tunin
g
for
Par
alle
lP
rogra
ms.
S0
8a.
1M
.K
ola
rIn
finit
esim
alsy
mm
etri
esof
wea
kly
pse
udoco
nvex
man
ifold
s.
S1
1a.
3C
.P
feif
er
Em
bed
din
gan
empir
ical
ly
bas
eddec
isio
nst
rate
gy
for
bac
kco
untr
ysk
iers
into
a
subje
ctiv
e...
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M1
:
Ind
ust
ria
la
nd
Ap
pli
ed
Ma
them
ati
cs
Ro
om
W2
05
Secti
on
S1
1:
Pro
ba
bil
ity,S
tati
stic
s
Ro
om
W2
06
Secti
on
S0
8:
Fu
ncti
on
al
An
aly
sis,
Rea
la
nd
Co
mp
lex
An
aly
sis
Ro
om
U4
06
Min
isy
mp
osi
um
M3
:
Pa
ra
llel
an
dD
istr
ibu
ted
Op
tim
iza
tio
na
nd
Sim
ula
tio
n
15:1
5
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
S0
1a.
3J.S
ch
ilh
an
The
tow
ersp
ectr
um
.
S0
1a.
2J.P
.S
ch
ürz
Str
ong
mea
sure
zero
sets
in
the
hig
her
Can
tor
spac
e.
M6
a.1
W.
Fa
ng
The
Ste
ep-B
ounce
Zet
aM
ap
inP
arab
oli
cC
atal
and.
S0
1a.
1M
.K
oelb
ing
Dis
trib
uti
vit
yof
forc
ing
noti
ons.
M6
a.2
A.R
.M
ille
rS
om
ere
sult
son
char
acte
rs
of
sym
met
ric
gro
ups.
Ro
om
U4
07
Min
isy
mp
osi
um
M6
:
En
um
era
tive
an
d
Alg
eb
ra
ic
Co
mb
ina
toric
s
Ro
om
U2
10
Ro
om
U4
05
Secti
on
S0
1:
Lo
gic
,S
et
Th
eo
ry,
an
d
Th
eo
reti
ca
lC
om
pu
ter
Scie
nce
Detailed Program of Sections and Minisymposia 33
Mon
day,
Sep
tem
ber
16,E
veni
ngS
essi
ons
17:1
5
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
S1
1b.1
W.
Wo
ess
Rec
urr
ence
of
2-d
imen
sional
queu
eing
pro
cess
es,
and
random
wal
kex
itti
mes
from
the.
..
M1
b.1
G.Z
an
gerl
Full
fiel
din
ver
sion
in
photo
acoust
icto
mogra
phy
wit
hvar
iable
sound
spee
d.
M1
b.2
B.
Ka
lten
ba
ch
er
Par
amet
erid
enti
fica
tion
and
unce
rtai
nty
quan
tifi
cati
on
in
stoch
asti
cst
ate
spac
e
model
s...
S1
1b.2
E.
Sö
nm
ez
Ran
dom
Wal
kon
the
Ran
dom
Connec
tion
Model
.
M3
b.1
K.R
hein
berg
er
Anal
yzi
ng
Auto
nom
ous
Dem
and
Sid
eA
lgori
thm
s
wit
hP
aral
lel
Com
puta
tion
Fra
mew
ork
s.
S0
8b.2
E.
Nig
sch
The
geo
met
riza
tion
of
the
theo
ryof
full
Colo
mbea
u
algeb
ras.
S0
8b.1
C.B
arg
etz
On
Gen
eric
Pro
per
ties
of
Nonex
pan
sive
Map
pin
gs.
M1
b.3
L.
Neu
ma
nn
Blo
ckco
ord
inat
edes
cent
met
hod
for
ill-
pose
d
pro
ble
ms.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M1
:
Ind
ust
ria
la
nd
Ap
pli
ed
Ma
them
ati
cs
Ro
om
W2
05
Secti
on
S1
1:
Pro
ba
bil
ity,S
tati
stic
s
Ro
om
W2
06
Secti
on
S0
8:
Fu
ncti
on
al
An
aly
sis,
Rea
la
nd
Co
mp
lex
An
aly
sis
Ro
om
U4
06
Min
isy
mp
osi
um
M3
:
Pa
ra
llel
an
dD
istr
ibu
ted
Op
tim
iza
tio
na
nd
Sim
ula
tio
n
17:1
5
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
M6
b.1
M.
J.S
ch
loss
er
Aw
eight-
dep
enden
t
inver
sion
stat
isti
can
d
Cat
alan
num
ber
s.
S0
1b.1
L.
Ha
lbeis
en
Form
endes
Ausw
ahla
xio
ms
inder
Rin
gth
eori
e.
S0
1b.2
V.T
orre
s-P
ere
zD
iam
onds,
gam
esan
d
card
inal
invar
iants
.
Ro
om
U4
07
Min
isy
mp
osi
um
M6
:
En
um
era
tive
an
d
Alg
eb
ra
ic
Co
mb
ina
toric
s
Ro
om
U2
10
Ro
om
U4
05
Secti
on
S0
1:
Lo
gic
,S
et
Th
eo
ry,
an
d
Th
eo
reti
ca
lC
om
pu
ter
Scie
nce
34 Detailed Program of Sections and Minisymposia
Tues
day,
Sep
tem
ber
17,M
orni
ngS
essi
ons
10:0
0
10:1
5
10:3
0
10:4
5
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
M1
c.3
P.S
ell
ars
Hyper
spec
tral
Imag
e
clas
sifi
cati
on
wit
hM
inim
al
Lea
rnin
g:
AG
raph
Bas
ed
Appro
ach.
S0
7a.
4A
.K
iese
nh
ofe
rS
mal
ldat
aglo
bal
regula
rity
for
hal
f-w
ave
map
sin
4
dim
ensi
ons.
S0
8c.
3F
.H
asl
ing
er
The
d-c
om
ple
xon
the
Seg
al-B
argm
ann
spac
e.
M1
c.2
D.
Ro
therm
el
Nonli
nea
rin
ver
sehea
t
tran
sfer
pro
ble
ms
in
model
ling
the
cooli
ng
pro
cess
of
hea
vy
pla
tes.
S0
7a.
1B
.S
ch
örk
hu
ber
Nonli
nea
ras
ym
pto
tic
stab
ilit
yof
hom
oth
etic
ally
shri
nkin
gY
ang-M
ills
soli
tons.
M1
c.1
M.
Lie
bre
ch
tP
roce
ssM
odel
for
the
Indust
rial
Pla
teL
evel
ing
Oper
atio
nw
ith
Spec
ial
Em
phas
ison
Onli
ne.
..
S0
7a.
2R
.B
ürg
er
Tw
o-l
ocu
scl
ines
mai
nta
ined
by
dif
fusi
on
and
reco
mbin
atio
nin
a
het
erogen
eous
envir
onm
ent.
S0
5a.
5Z
.V
irk
Det
ecti
ng
length
sof
geo
des
ics
wit
h
hig
her
-dim
ensi
onal
per
sist
ence
.
S0
7a.
3S
.B
latt
The
neg
ativ
eL
2gra
die
nt
flow
for
p-e
last
icen
ergie
s.
M1
c.5
A.M
aie
rK
now
nO
per
ator
Lea
rnin
gU
An
Appro
ach
tounit
e
Physi
cs,S
ignal
Pro
cess
ing,
and
Mac
hin
e...
M1
c.4
J.S
ch
wa
bR
egula
rizi
ng
Net
work
sfo
r
Solv
ing
Inver
seP
roble
ms.
S0
8c.
4K
.G
roech
en
igS
ampli
ng,
Mar
cinkie
wic
z-Z
ygm
und
sets
,ap
pro
xim
atio
n,
and
quad
ratu
reru
les.
S0
8c.
1M
.D
ym
on
dH
ighly
irre
gula
rse
par
ated
net
s..
S0
8c.
2I.
Ko
ssov
skiy
Cla
ssifi
cati
on
of
hom
ogen
eous
stri
ctly
pse
udoco
nvex
hyper
surf
aces
inC
3.
S0
5a.
3A
.S
.R
oll
eS
pac
esof
clust
erin
gs.
S0
5a.
1R
.K
wit
tD
eep
Hom
olo
gic
al
Lea
rnin
g.
S0
5a.
4A
.N
ikit
en
ko
Sto
chas
tic
appli
cati
ons
of
dis
cret
eto
polo
gy.
S0
5a.
2D
.E
ga
sS
an
tan
der
Topolo
gic
alex
plo
rati
ons
in
neu
rosc
ience
.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M1
:
Ind
ust
ria
la
nd
Ap
pli
ed
Ma
them
ati
cs
Ro
om
W2
05
Secti
on
S0
7:
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
an
dC
alc
ulu
s
of
Va
ria
tio
ns
Ro
om
W2
06
Secti
on
S0
8:
Fu
ncti
on
al
An
aly
sis,
Rea
la
nd
Co
mp
lex
An
aly
sis
Ro
om
U4
06
Secti
on
S0
5:
Co
mp
uta
tio
na
l
Geo
metr
ya
nd
To
po
log
y
10:0
0
10:1
5
10:3
0
10:4
5
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
S1
2.2
S.D
esm
ett
reS
ever
ity
Model
ing
of
Extr
eme
Insu
rance
Cla
ims
for
Tar
iffi
cati
on.
S1
2.1
J.S
ass
Uti
lity
Max
imiz
ing
Str
ateg
ies
under
Incr
easi
ng
Model
Unce
rtai
nty
ina
Mult
ivar
iate
Bla
ck..
.
S1
2.3
J.A
.S
trin
iO
pti
mal
div
iden
ds
and
fundin
gin
risk
theo
ry.
S1
2.4
S.T
ho
nh
au
ser
Opti
mal
rein
sura
nce
for
expec
ted
pen
alty
funct
ions
inri
skth
eory
.
S0
1c.
4W
.W
oh
ofs
ky
Car
din
alch
arac
teri
stic
san
d
the
Bore
lC
onje
cture
.
M6
c.1
M.
Ka
uers
Mak
ing
Man
yM
ore
Mat
rix
Mult
ipli
cati
on
Met
hods.
S0
1c.
2S
.S
ch
um
ach
er
The
rela
tion
bet
wee
ntw
o
wea
kch
oic
epri
nci
ple
s.
M6
c.2
B.
Stu
fler
Loca
lco
nver
gen
ceof
random
pla
nar
gra
phs.
M6
c.3
E.
Y.J
inE
xac
tan
das
ym
pto
tic
enum
erat
ion
of
asce
nt
sequen
ces
and
Fis
hburn
mat
rice
s.
S0
1c.
3V
.F
isch
er
Indep
enden
cean
dal
most
dis
join
tnes
s.
S0
1c.
1M
.M
üll
er
Auto
mat
ing
Res
olu
tion
is
NP
-Har
d.
Ro
om
U4
07
Min
isy
mp
osi
um
M6
:
En
um
era
tive
an
d
Alg
eb
ra
ic
Co
mb
ina
toric
s
Ro
om
U2
10
Secti
on
S1
2:
Fin
an
cia
la
nd
Actu
aria
l
Ma
them
ati
cs
Ro
om
U4
05
Secti
on
S0
1:
Lo
gic
,S
et
Th
eo
ry,
an
d
Th
eo
reti
ca
lC
om
pu
ter
Scie
nce
Detailed Program of Sections and Minisymposia 35
Tues
day,
Sep
tem
ber
17,A
fter
noon
Ses
sion
s
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
17:0
0
S0
7b.3
L.
Co
ssett
iU
niq
ue
Conti
nuat
ion
for
the
Zak
har
ov
Kuzn
etso
v
equat
ion.
S0
5b.2
L.
L.
Cris
tea
On
frac
tal
geo
met
ric
and
topolo
gic
pro
per
ties
of
tria
ngula
rla
byri
nth
frac
tals
.
S0
9a.
3T
.C
ati
na
sO
nso
me
inte
rpola
tion
oper
ators
on
tria
ngle
sw
ith
curv
edsi
des
.
S0
7b.1
I.G
log
icT
hre
shold
for
blo
wup
for
the
super
crit
ical
cubic
wav
e
equat
ion.
S0
9a.
2R
.E
berle
Vib
rati
on
of
abea
munder
stru
ctura
lra
ndom
nes
s.
S0
7b.2
N.
Vo
rd
ero
berm
eie
rO
nth
ean
alyti
city
of
crit
ical
poin
tsof
the
Möbiu
sen
ergy.
S0
8d
.1S
.H
all
er
The
hea
tker
nel
expan
sion
of
Rock
land
dif
fere
nti
al
oper
ators
and
appli
cati
ons
to
gen
eric
...
S0
5b.3
G.L
eo
ba
ch
er
Nec
essa
ryan
dsu
ffici
ent
condit
ions
for
the
loca
l
uniq
ue
nea
rest
poin
t
pro
per
tyof
are
al..
.
S0
9a.
1O
.S
tein
ba
ch
Spac
e-T
ime
Fin
ite
Ele
men
t
Met
hods.
S0
5b.1
J.L
ian
gT
opolo
gic
alS
truct
ure
sof
Pro
bab
ilis
tic
Lan
dsc
ape
of
Sto
chas
tic
Rea
ctio
n
Net
work
s.
S0
8d
.2U
.M
.H
an
un
gA
role
of
Har
nac
kex
tensi
on
inin
tegra
ting
funct
ions
over
arbit
rary
subse
ts.
S0
8d
.3U
.G
ru
pel
Inte
rsec
tions
wit
hra
ndom
geo
des
ics
inhig
h
dim
ensi
ons.
Ro
om
W2
11
/12
Secti
on
S0
9:
Nu
meric
al
An
aly
sis,
Scie
nti
fic
Co
mp
uti
ng
Ro
om
W2
05
Secti
on
S0
7:
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
an
dC
alc
ulu
s
of
Va
ria
tio
ns
Ro
om
W2
06
Secti
on
S0
8:
Fu
ncti
on
al
An
aly
sis,
Rea
la
nd
Co
mp
lex
An
aly
sis
Ro
om
U4
06
Secti
on
S0
5:
Co
mp
uta
tio
na
l
Geo
metr
ya
nd
To
po
log
y
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
17:0
0
S0
2a.
1A
.P
an
ho
lzer
Lab
el-q
uan
titi
esin
tree
san
d
map
pin
gs.
S0
2a.
2D
.K
ren
nO
pti
mal
Mult
i-P
ivot
Quic
kso
rtan
dit
s
Asy
mpto
tic
Anal
ysi
s.
S0
3a.
1K
.D
est
ag
no
lC
ounti
ng
poin
tsof
giv
en
deg
ree
and
bounded
hei
ght
via
the
hei
ght
zeta
funct
ion.
S0
3a.
2N
.R
om
eO
nS
erre
’sP
roble
mfo
r
Conic
s.
S0
2a.
3M
.W
ibm
er
Gal
ois
gro
ups
of
dif
fere
nti
al
equat
ions.
Ro
om
U4
07
Secti
on
S0
2:
Alg
eb
ra
,D
iscre
te
Ma
them
ati
cs
Ro
om
U2
10
Ro
om
U4
05
Secti
on
S0
3:
Nu
mb
er
Th
eo
ry
36 Detailed Program of Sections and Minisymposia
Tues
day,
Sep
tem
ber
17,E
veni
ngS
essi
ons
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
19:0
0
S0
9b.3
A.K
.Y
ad
av
Eff
ect
of
Imped
ance
on
Refl
ecti
on
of
Pla
ne
Wav
esin
aR
ota
ting
Mag
net
o-t
her
moel
asti
c...
S0
9b.2
S.
Sin
gh
Anew
appro
ach
of
oper
atio
nal
mat
rice
sfo
r
hyper
boli
cpar
tial
dif
fere
nti
aleq
uat
ions.
S0
9b.1
M.
Mele
nk
anad
apti
ve
algori
thm
for
the
frac
tional
Lap
laci
an.
S0
7c.
3A
.T
um
an
ya
nF
redholm
pro
per
tyof
regula
rhypoel
lipti
c
oper
ators
inm
ult
ianis
otr
opic
Sobole
vsp
aces
.
S0
7c.
1L
.F
orc
ell
aIn
terp
ola
tion
theo
ryan
d
regula
rity
for
inco
mpre
ssib
le
fluid
model
s.
S0
5c.
1A
.B
row
nC
onver
gen
cean
dS
tabil
ity
of
Ran
dom
Map
per
.
S0
5c.
2U
.B
au
er
Dual
itie
san
dcl
eari
ng
for
imag
eper
sist
ence
.
S0
7c.
2D
.S
.W
all
au
ch
Sta
ble
blo
wup
for
a
nonli
nea
rK
lein
-Gord
on
equat
ion.
Ro
om
W2
11
/12
Secti
on
S0
9:
Nu
meric
al
An
aly
sis,
Scie
nti
fic
Co
mp
uti
ng
Ro
om
W2
05
Secti
on
S0
7:
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
an
dC
alc
ulu
s
of
Va
ria
tio
ns
Ro
om
W2
06
Ro
om
U4
06
Secti
on
S0
5:
Co
mp
uta
tio
na
l
Geo
metr
ya
nd
To
po
log
y
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
19:0
0
S0
3b.2
S.
Ya
ma
gis
hi
Solv
ing
equat
ions
inm
any
var
iable
sin
pri
mes
.
S0
4.1
H.M
.A
l-Z
ou
bi
AN
CH
OR
RIN
GS
OF
FIN
ITE
TY
PE
GA
US
S
MA
PIN
TH
E
EU
CL
IDE
AN
3-S
PA
CE
.
S0
3b.1
C.F
rei
Ari
thm
etic
pro
gre
ssio
ns
in
bin
ary
quad
rati
cfo
rms
and
norm
form
s.
Ro
om
U4
07
Ro
om
U2
10
Ro
om
U4
05
Secti
on
S0
3:
Nu
mb
er
Th
eo
ry
Detailed Program of Sections and Minisymposia 37
Wed
nesd
ay,S
epte
mbe
r18
,Mor
ning
Ses
sion
s
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
S1
0.1
I.A
.R
.M
og
hra
bi
AM
ult
i-st
epH
ybri
d
Conju
gat
eG
radie
nt
Met
hod.
S0
6.2
Y.
Ha
lim
On
asy
stem
of
dif
fere
nce
equat
ions
of
seco
nd
ord
er
solv
edin
acl
ose
dfr
om
.
S0
9c.
1V
.D
evi
An
oper
atio
nal
mat
rix
appro
ach
for
two-d
imen
sional
hyper
boli
c
tele
gra
ph
equat
ion.
S0
9c.
2R
.K
.M
au
ry
aM
ult
iste
pfi
nit
edif
fere
nce
schem
efo
rel
ectr
om
agnet
ic
wav
esm
odel
aris
ing
from
...
M5
a.1
M.
Ha
ltm
eie
rN
ET
T:
Solv
ing
Inver
se
Pro
ble
ms
wit
hD
eep
Neu
ral
Net
work
s.
M5
a.2
D.
Ob
ma
nn
Spar
sesy
nth
esis
regula
riza
tion
usi
ng
dee
p
neu
ral
net
work
s.
S0
6.3
C.E
.A
rro
ud
Opti
mal
ity
condit
ions
of
More
ausw
eepin
gpro
cess
.
S0
6.1
P.S
zm
oly
an
Per
iodic
solu
tions
and
and
swit
chin
gbeh
avio
rin
a
model
of
bip
ola
rdis
ord
er.
M5
a.3
R.R
am
lau
Sin
gula
rV
alue
Dec
om
posi
tion
for
Atm
osp
her
icT
om
ogra
phy.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M5
:
Inverse
Pro
ble
ms
for
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
Ro
om
W2
05
Secti
on
S0
6:
Ord
ina
ry
Dif
fere
nti
al
Eq
ua
tio
ns
an
d
Dy
na
mic
al
Sy
stem
s
Ro
om
W2
06
Secti
on
S0
9:
Nu
meric
al
An
aly
sis,
Scie
nti
fic
Co
mp
uti
ng
Ro
om
U4
06
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
Ro
om
U4
07
Ro
om
U2
10
Ro
om
U4
05
38 Detailed Program of Sections and Minisymposia
Thur
sday
,Sep
tem
ber
19,M
orni
ngS
essi
ons
10:0
0
10:1
5
10:3
0
10:4
5
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
M4
a.5
A.S
ak
hn
ov
ich
Dis
cret
eD
irac
syst
ems:
spec
tral
and
scat
teri
ng
theo
ries
and
Ver
blu
nsk
y-t
ype
coef
fici
ents
.
M5
b.2
F.H
ett
lich
Iden
tifi
cati
on
and
Des
ign
of
EM
-Chir
alS
catt
erin
g
Obst
acle
s.
M5
b.3
A.
Ag
alt
sov
Uniq
uen
ess
and
reco
nst
ruct
ion
ina
pas
sive
inver
sepro
ble
mof
hel
iose
ism
olo
gy.
M5
b.1
W.
Ru
nd
ell
Det
erm
inin
gnonli
nea
rte
rms
ina
reac
tion-d
iffu
sion
syst
em.
M5
b.5
G.M
erc
ier
On
conver
gen
ceof
Tota
l
Var
iati
on
regula
rize
din
ver
se
pro
ble
ms.
M4
a.1
T.K
ap
pele
rO
nth
ein
tegra
bil
ity
of
the
Ben
jam
in-O
no
equat
ion
wit
hper
iodic
boundar
y
condit
ions.
M4
a.3
A.K
ost
en
ko
On
the
abso
lute
ly
conti
nuous
spec
trum
of
gen
eral
ized
indefi
nit
e
stri
ngs.
M5
b.4
A.
Wa
ldP
aram
eter
iden
tifi
cati
on
for
the
Lan
dau
-Lif
shit
z-G
ilber
t
equat
ion
inM
agnet
ic
Par
ticl
e...
M4
a.4
R.D
on
nin
ger
Str
ichar
tzes
tim
ates
for
the
one-
dim
ensi
onal
wav
e
equat
ion.
M4
a.2
D.
Len
zU
nif
orm
ity
of
cocy
cles
and
spec
tra
of
Sch
reie
rgra
phs.
M2
a.5
A.K
ra
tsio
sU
niv
ersa
lA
ppro
xim
atio
n
Theo
rem
s.
M2
a.4
J.T
eic
hm
an
nL
earn
ing
stoch
asti
c
dynam
ics
by
random
signat
ure
sw
ith
appli
cati
ons
tom
athem
atic
al..
.
M2
a.3
L.
Go
no
nE
rror
Bounds
for
Ran
dom
Rec
urr
ent
Neu
ral
Net
work
s
and
Gen
eral
Res
ervoir
Com
puti
ng
Syst
ems.
M2
a.1
J.P
.O
rte
ga
Dynam
ican
dC
ontr
ol
Theo
reti
cal
Asp
ects
of
Res
ervoir
Com
puti
ng.
M2
a.2
L.
Grig
oryev
aF
ore
cast
ing
of
Rea
lize
d
(Co)V
aria
nce
sw
ith
Res
ervoir
Com
puti
ng.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M5
:
Inverse
Pro
ble
ms
for
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
Ro
om
W2
05
Min
isy
mp
osi
um
M2
:
Ma
them
ati
ca
lF
ina
nce
inth
ea
ge
of
Ma
ch
ine
Lea
rnin
g
Ro
om
W2
06
Min
isy
mp
osi
um
M4
:
Sp
ectr
al
Th
eo
ry
Ro
om
U4
06
10:0
0
10:1
5
10:3
0
10:4
5
11:0
0
11:1
5
11:3
0
11:4
5
12:0
0
12:1
5
12:3
0
Ro
om
U4
07
Ro
om
U2
10
Ro
om
U4
05
Detailed Program of Sections and Minisymposia 39
Thur
sday
,Sep
tem
ber
19,A
fter
noon
Ses
sion
s
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
17:0
0
M4
b.1
G.D
.R
ay
kov
Thre
shold
Sin
gula
riti
esof
the
Spec
tral
Shif
tF
unct
ion
for
Geo
met
ric
Per
turb
atio
ns
of
a...
M5
c.3
R.K
lein
Reg
ula
rizi
ng
sequen
tial
subsp
ace
opti
miz
atio
nfo
r
the
iden
tifi
cati
on
of
the
store
den
ergy..
.
M5
c.2
F.R
om
ero
Hin
ric
hse
nD
ynam
ical
super
-res
olu
tion
wit
hap
pli
cati
ons
toU
ltra
fast
ult
raso
und..
.
M4
b.3
M.
Pio
rk
ow
ski
Rie
man
n-H
ilber
tap
pro
ach
toas
ym
pto
tic
anal
ysi
s.
M5
c.1
I.P
iotr
ow
ska
-K
urc
zew
ski
Gen
eral
ized
tole
rance
regula
riza
tion.
M2
b.3
W.
Kh
osr
aw
iA
Neu
ral
Net
work
Appro
ach
toL
oca
lS
toch
asti
cV
ola
tili
ty
Cal
ibra
tion.
M4
b.2
J.M
ich
or
Rar
efac
tion
and
shock
wav
esfo
rth
eT
oda
equat
ion.
M2
b.1
H.S
.W
utt
eR
andom
ized
shal
low
neu
ral
net
work
san
dth
eir
use
in
under
stan
din
ggra
die
nt
des
cent.
M2
b.2
C.
Cu
ch
iero
Model
ing
rough
covar
iance
pro
cess
es.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M5
:
Inverse
Pro
ble
ms
for
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
Ro
om
W2
05
Min
isy
mp
osi
um
M2
:
Ma
them
ati
ca
lF
ina
nce
inth
ea
ge
of
Ma
ch
ine
Lea
rnin
g
Ro
om
W2
06
Min
isy
mp
osi
um
M4
:
Sp
ectr
al
Th
eo
ry
Ro
om
U4
06
15:3
0
15:4
5
16:0
0
16:1
5
16:3
0
16:4
5
17:0
0
Ro
om
U4
07
Ro
om
U2
10
Ro
om
U4
05
40 Detailed Program of Sections and Minisymposia
Thur
sday
,Sep
tem
ber
19,E
veni
ngS
essi
ons
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
19:0
0
M4
c.1
R.S
ou
saP
roduct
form
ula
san
d
convolu
tions
for
solu
tions
of
Stu
rm-L
iouvil
leeq
uat
ions.
M5
d.1
S.
Eb
erle
Solv
ing
the
inver
sepro
ble
m
of
linea
rel
asti
city
wit
h
monoto
nic
ity
met
hods.
M4
c.2
Z.
Ra
oO
pti
mal
Blo
wup
Sta
bil
ity
for
the
Ener
gy
Cri
tica
lW
ave
Equat
ion.
M2
c.3
S.R
igg
er
Inte
ract
ing
Par
ticl
eS
yst
ems,
Def
ault
Cas
cades
and
the
M1-t
opolo
gy.
M2
c.1
T.K
ra
bic
hle
rD
eep
AL
M.
M2
c.2
S.S
va
luto
-Ferro
Infi
nit
edim
ensi
onal
poly
nom
ial
jum
p-d
iffu
sions.
Ro
om
W2
11
/12
Min
isy
mp
osi
um
M5
:
Inverse
Pro
ble
ms
for
Pa
rti
al
Dif
fere
nti
al
Eq
ua
tio
ns
Ro
om
W2
05
Min
isy
mp
osi
um
M2
:
Ma
them
ati
ca
lF
ina
nce
inth
ea
ge
of
Ma
ch
ine
Lea
rnin
g
Ro
om
W2
06
Min
isy
mp
osi
um
M4
:
Sp
ectr
al
Th
eo
ry
Ro
om
U4
06
17:3
0
17:4
5
18:0
0
18:1
5
18:3
0
18:4
5
19:0
0
Ro
om
U4
07
Ro
om
U2
10
Ro
om
U4
05
ABSTRACTS OF PLENARYLECTURES, MINISYMPOSIAAND SECTIONS
41
42 ABSTRACTS
Plenary Speakers
Efstathia Bura (TU Wien)Christopher Frei (University of Manchester, ÖMG Award Winner)Fritz Gesztesy (Baylor University)Marlis Hochbruck (KIT Karlsruhe)Josef Hofbauer (University of Wien)Özlem Imamoglu (ETH Zürich)Helmut Pottmann (KAUST and TU Wien)Wilhelm Schlag (Yale University)Susanna Terracini (University of Turin)
44 Plenary Speakers
P.5
TUE14:00
I15:00
W211/12
Sufficient Dimension Reduction in high dimensional regression and classifica-tionEfstathia Bura (Vienna University of Technology)
This talk will focus on model-based sufficient dimension reduction in regressionsand classifications with predictor distributions in the elliptically contoured and ex-ponential families. Model-based SDR stems from the connection between sufficientstatistics in the inverse regression and the sufficient reduction in the forward regres-sion. This leads to the exhaustive identification of the sufficient reductions, bothlinear and nonlinear, in the elliptically contoured and exponential families of dis-tributions, which model most of the continuous, discrete and mixed predictors inhigh dimensional problems. The work associated with these developments will bepresented.
P.7
WED09:30
I10:30
W211/12
Average bounds for the `-torsion in class groups of number fieldsChristopher Frei? (University of Manchester)Martin Widmer (Royal Holloway, University of London)
The arithmetic of an algebraic number field K is determined in large parts by itsclass group ClK . For example, the ring of integers of K admits unique factorisationinto irreducibles if and only if ClK is trivial. While it is known from classical alge-braic number theory that ClK is always a finite abelian group, and the distribution ofthese groups is predicted by precise heuristics, precise provable information abouttheir structure remains quite elusive. To emphasise our lack of knowledge, the ques-tion whether the class group is trivial for infinitely many number fields remains stillopen.
In this talk, we introduce class groups and survey classical and recent results (con-ditional and unconditional) bounding the cardinality of their `-torsion subgroups, fora natural number `. In the remaining time, we discuss joint work with Martin Wid-mer on average bounds for this `-torsion in certain families of number fields, andapplications thereof.
P.9
THU14:00
I15:00
W211/12
The Birman sequence of integral inequalities and some of its refinementsFritz Gesztesy (Department of Mathematics, Baylor University, Waco, Texas)
We review recent progress on the Birman sequence of integral inequalities in arbi-trary dimension (the first two elements in this sequence being the celebrated inequal-ities of Hardy and Rellich).
Refinements of these inequalities that we will discuss include power weights aswell as an arbitrary number of additional logarithmically weaker singular terms. Inaddition, we show that powers of the Laplacian can be replaced by the correspondingpowers of the radial Laplacian.
This is based on joint work with L. L. Littlejohn, I. Michael, and M. M. H. Pang.
Plenary Speakers 45
P.3
MON13:45
I14:45
W211/12
On leapfrog-Chebyshev schemesMarlis Hochbruck (Karlsruhe Institute of Technology)
In this talk we consider variants of the leapfrog method for second order differ-ential equations. In numerous situations the strict CFL condition of the standardleapfrog method is the main bottleneck that thwarts its performance. Based onChebyshev polynomials new methods have been constructed that exhibit a muchweaker CFL condition than the leapfrog method.
We will analyze the stability and the long-time behavior of leapfrog-Chebyshevmethods in two-step formulation. This analysis indicates that on should modify theschemes proposed in the literature in two ways to improve their qualitative behavior:
For linear problems, we propose to use special starting values required for a two-step method instead of the standard choice obtained from Taylor approximation. Forsemilinear problems, we propose to introduce a stabilization parameter as it has beendone for Runge-Kutta-Chebyshev methods before. For the stabilized methods weprove that they guarantee stability for a large class of problems.
While these results hold for quite general polynomials, we show that the polyno-mials used in the stabilized leapfrog Chebyshev methods satisfy all the necessaryconditions. In fact, all constants in the error analysis can be given explicitly.
The talk concludes with numerical examples illustrating our theoretical findings.This is joint work with Constantin Carle and Andreas Sturm, Karlsruhe Institute
of Technology, funded by CRC 1173.
P.6
WED08:30
I09:30
W211/12
Permanence of biological systemsJosef Hofbauer (Universität Wien)
Given a mathematical model of a biological community [5, 6] (or a chemical re-action network [2]), can one find out whether the whole community will survive? Isthere is a global attractor in which all species/types of the system coexist? If the statespace is the nonnegative orthant Rn
+ this means that its boundary (together with ∞) isa repeller for the dynamics. Important tools are average Lyapunov functions [1, 3, 5]and Lyapunov exponents [4]. I will give a survey of the methods and results obtainedin the last decades about this problem.
[1] M. Benaïm, Stochastic persistence, https://arxiv.org/abs/1806.08450[2] B. Boros, J. H., Permanence of weakly reversible mass-action systems with a single
linkage class, https://arxiv.org/abs/1903.03071[3] B. Garay, J. H., Robust permanence for ecological differential equations, minimax, and
discretizations, SIAM J. Math. Anal. 34 (2003) 1007–1039.[4] J. H., S. Schreiber, To persist or not to persist, Nonlinearity 17 (2004) 1393–1406.[5] J. H., K. Sigmund, Evolutionary Games and Population Dynamics, Cambridge Univer-
sity Press (1998).[6] H. L. Smith, H. R. Thieme, Dynamical Systems and Population Persistence,
Amer. Math. Soc., 2011.
46 Plenary Speakers
P.4
TUE08:30
I09:30
W211/12
Binary Quadratic Forms, Quadratic fields and Modular formsÖzlem Imamoglu (ETH Zürich)
There are numerous connections between binary quadratic forms, quadratic fieldsand modular forms. One of the most beautiful is provided by the theory of singularmoduli, which are the values of the Klein’s j-invariant
j(τ) = q−1 +744+196884q+21493760q2 + · · · (q = e2πiτ)
at imaginary quadratic irrationalities. In this talk I will give an overview of some ofthe old and new results and various applications of the connections between quadraticforms and modular forms.
P.2
MON11:00
I12:00
W211/12
Discrete geometric structures motivated by architecture and computational de-signHelmut Pottmann (KAUST and TU Wien)
Discrete differential geometry is situated at the border between differential anddiscrete geometry and aims to develop discrete equivalents of notions and methodsfrom smooth surface theory [1]. The current interest in discrete differential geometryis not only rooted in its importance in pure mathematics, but also in its relevance forother fields, including computer graphics and architectural geometry.
This talk will present recent results in discrete differential geometry whose devel-opment has been motivated by practical requirements on the fabrication of freeformshapes in modern architecture [3]. We will focus on various discretizations of princi-pal curvature parameterizations [2]. The geometric properties of these principal netsfacilitate the construction of architectural freeform skins from flat, nearly rectangu-lar panels, associated with optimized support structures. Principal nets also appearin material-minimizing structures [4], and they are in a certain sense the smoothestpolyhedral surfaces which approximate a given smooth surface [5]. Non-smoothrepresentatives of certain principal nets are related to curved folding patterns andorigami.
[1] A. Bobenko, Y. Suris, Discrete differential geometry: Integrable Structure, AmericanMath. Soc., 2009.
[2] A. Bobenko, H. Pottmann, J. Wallner, A curvature theory for discrete surfaces based onmesh parallelity, Math. Annalen, 348 (2010), 1–24.
[3] H. Pottmann, M. Eigensatz, A. Vaxman, J. Wallner, Architectural geometry, Computersand Graphics, 47 (2015) 145–164.
[4] D. Pellis, M. Kilian, J. Wallner, H. Pottmann, Material-minimizing forms and structures,ACM. Trans. Graphics 36(6) (2017), 173:1–173:12.
[5] D. Pellis, M. Kilian, F. Dellinger, J. Wallner, H. Pottmann, Visual smoothness of polyhe-dral surfaces, ACM. Trans. Graphics 38(4) (2019), 31:1–31:11.
Plenary Speakers 47
P.1
MON09:30
I10:30
W211/12
On no-return theorems in nonlinear dispersive PDEsWilhelm Schlag (Yale University)
In 2010 Nakanishi and the speaker introduced a "one-pass theorem" which pre-cludes solutions to nonlinear wave equations starting from a small neighborhood ofa ground state soliton and which return to another such neighborhood after somefinite time. Such statements are non-perturbative and require global arguments. Inthe original setting invariant manifolds played an essential role, especially the one-dimensional unstable manifold associated with the steady-state solution. I will re-view these results, and discuss some of their ramifications over the past ten years.
48 Plenary Speakers
P.8
THU08:30
I09:30
W211/12
Spiralling and other solutions in limiting profiles of competition-diffusionsystemsSusanna Terracini (Università di Torino)
Reaction-diffusion systems with strong interaction terms appear in many multi-species physical problems as well as in population dynamics. The qualitative prop-erties of the solutions and their limiting profiles in different regimes have been at thecenter of the community’s attention in recent years. A prototypical example is thesystem of equations
−∆u+a1u = b1|u|p+q−2u+ cp|u|p−2|v|qu,−∆v+a2v = b2|v|p+q−2v+ cq|u|p|v|q−2v
in a domain Ω⊂ RN which appears, for example, when looking for solitary wavesolutions for Bose-Einstein condensates of two different hyperfine states which over-lap in space. The sign of bi reflects the interaction of the particles within each singlestate. If bi is positive, the self interaction is attractive (focusing problems). Thesign of c, on the other hand, reflects the interaction of particles in different states.This interaction is attractive if c > 0 and repulsive if c < 0. If the condensates repel,they eventually separate spatially giving rise to a free boundary. Similar phenomenaoccurs for many species systems. As a model problem, we consider the system ofstationary equations:
−∆ui = fi(ui)−βui ∑ j 6=i gi j(u j)
ui > 0 .
The cases gi j(s) = βi js (Lotka-Volterra competitive interactions) and gi j(s) = βi js2
(gradient system for Gross-Pitaevskii energies) are of particular interest in the appli-cations to population dynamics and theoretical physics respectively.
Phase separation and has been described in the recent literature, both physical andmathematical. Relevant connections have been established with optimal partitionproblems involving spectral functionals. The classification of entire solutions andthe geometric aspects of phase separation are of fundamental importance as well.We intend to focus on the most recent developments of the theory in connection withproblems featuring anomalous diffusions, non-local and non symmetric interactions.
Minisymposium M1Industrial and Applied Mathematics
Markus Haltmeier (Universität Innsbruck)
50 M1: Industrial and Applied Mathematics
M1a.1
MON15:15
I15:45
W211/12
Efficient solving of a time-dependent interface problem
Christoph Erath (Technische Universität Darmstadt)
Modeling has been identified as an important tool to simulate phenomena and
processes in natural sciences and engineering, which are often described by partial
differential equations. Appropriate and effective numerical methods play a crucial
role to solve these equations.
We consider a parabolic-elliptic interface problem in an unbounded domain. An ap-
plication is, for instance, the modeling of eddy currents in the magneto-quasistatic
regime. Due to different physical properties of the solution in different parts of the
domain it makes sense to consider couplings of different methods to get the best pos-
sible numerical approximation.
Therefore, we introduce the coupling of the Finite Element Method and the Bound-
ary Element Method to approximate such a solution in space without truncation of
the unbounded domain [1]. This leads to a semi-discrete solution. For the sub-
sequent time discretization we either use the classical backward Euler scheme or
a variant of it to get a fully-discrete system. The model problem and its numeri-
cal discretization are well-posed also for Lipschitz interfaces [2]. Even more, we
prove quasi-optimality, i.e., the best possible approximation of the solution, of our
semi-discrete and fully-discrete solution under minimal regularity assumptions in
the natural energy norm [2]. We remark that our analysis does not use duality argu-
ments and corresponding estimates for an elliptic projection which are not available
for our nonsymmetric coupling method. Instead, we use estimates in appropriate en-
ergy norms. Moreover, we discuss the extension of our model problem to a problem
arising in fluid mechanics. However, since the conservation of fluxes is mandatory
for such applications, we replace the Finite Element Method by the Finite Volume
Method in our coupling approach [3]. Numerical examples illustrate the predicted
(optimal) convergence rates and underline the potential for practical applications.
[1] R. C. MacCamy and M. Suri, A time-dependent interface problem for two-dimensional
eddy currents, Quart. Appl. Math., 44 (1987) 675–690.
[2] H. Egger, C. Erath, and R. Schorr, On the nonsymmetric coupling method for parabolic-
elliptic interface problems, SIAM J. Numer. Anal. 56(6) (2018) 3510–3533.
[3] C. Erath and R. Schorr, Stable non-symmetric coupling of the finite volume method
and the boundary element method for convection-domninated parabolic-elliptic inter-
face problems, Comput. Methods Appl. Math. published online (2019).
M1: Industrial and Applied Mathematics 51
M1a.2
MON15:45
I16:15
W211/12
Using extended Kalman Filters for solving non-linear inverse problemsKlaus Frick? (Interstaatliche Hochschule für Technik Buchs NTB)Silvano Balemi (Zumbach Electronic AG)
Since decades the Kalman filter approach is widely used in numerous fields of ap-plications ranging from inertial navigation to econometrics. Extended Kalman filtersare used to estimate a time dependent state variable xk ∈ X from noisy observationsyk ∈ Y according to the general non-linear dynamic model
xk+1 = f (xk)+ vk
yk = h(xk)+wk
Here f : X → X defines the dynamics of the system, h : X → Y is the measurementmapping and vk and wk are noise terms. In this talk we consider the case when themeasurement mapping is ill-posed and stress that in this case the Extended Kalmanfilter acts as a regularization method. We illustrate the performance of the methodby means of a real-world example from industrial inline metrology, where the cross-section of moving tubular products such as steel-bars or cables is reconstructed athigh precision from incomplete and noisy measurements.
M1a.3
MON16:15
I16:45
W211/12
Iterative methods for PAT in dissipative mediaRichard Kowar? (Department of Mathematics, University of Innsbruck)Markus Haltmeier (Department of Mathematics, University of Innsbruck)Linh V. Nguyen (Department of Mathematics, University of Idaho)
We present a convergence analysis of iterative methods for PAT in attenuatingmedium. An explicit expression for the adjoint problem is derived, which can ef-ficiently be implemented. In contrast to time reversal, the adjoint wave equationis damping and, thus has a stable solution. Our numerical results demonstrate effi-ciency, stability and accuracy of the iterative method in various situations includingthe limited view case.
52 M1: Industrial and Applied Mathematics
M1b.1
MON17:15
I17:45
W211/12
Full field inversion in photoacoustic tomography with variable sound speedGerhard Zangerl? (Department of Mathematics, University Innsbruck)Markus Haltmeier (Department of Mathematics, University Innsbruck)Linh V. Nguyen (Department of Mathematics, University Idaho)Robert Nuster (Department of Physics, University Graz)
Photoacoustic tomography (PAT) is a hybrid imaging technique, combining highcontrast and resolution of optical and ultrasound tomography. Typically, data ac-quisition in PAT requires to record acoustic pressure that is emitted by an object bydetectors surrounding it, which is typically time consuming. To address this problem,we consider a faster novel measurement setup for data acquisition in PAT where datain the form of projections of the full 3D acoustic pressure distribution at a certaintime instant are collected. Existing imaging algorithms for this kind of data assumeconstant speed of sound. This assumption is not always met in practice and thusleads to erroneous reconstructions. We present a two-step reconstruction methodfor full field detection photoacoustic tomography that takes variable speed of soundinto account. In the first step, by applying the inverse Radon transform, the pressuredistribution at the measurement time is reconstructed point-wise from the projectiondata. In the second step, one solves a final time wave inversion problem where theinitial pressure distribution is recovered from the known pressure distribution at themeasurement time. For the latter problem, we derive an iterative solution approach,compute the required adjoint operator, and show its uniqueness and stability.
This talk is based on [1].
[1] G. Zangerl, M. Haltmeier, L. V. Nguyen, and R. Nuster 2019. Full field inversion inphotoacoustic tomography with variable sound speed. Appl. Sci. 9 (8), pp. 1563.
M1: Industrial and Applied Mathematics 53
M1b.2
MON17:45
I18:15
W211/12
Parameter identification and uncertainty quantification in stochastic statespace models and its application to texture analysisBarbara Kaltenbacher? (Alpen-Adrai-Universität Klagenfurt)Barbara Pedretscher (Kompetenzzentrum Automobil- & Industrie-Elektronik Villach)Olivia Bluder (Kompetenzzentrum Automobil- & Industrie-Elektronik Villach)
In this talk, a computational framework, which enables efficient and robust param-eter identification, as well as uncertainty quantification in state space models basedon Itô stochastic processes, is presented. For optimization, a Maximum Likelihoodapproach based on the system’s corresponding Fokker-Planck equation is followed.Gradient information is included by means of an adjoint approach, which is based onthe Lagrangian of the optimization problem. To quantify the uncertainty of the Max-imum a Posteriori estimates of the model parameters, a Bayesian inference approachbased on Markov Chain Monte Carlo simulations, as well as profile likelihoods areimplemented and compared in terms of runtime and accuracy. The framework is ap-plied to experimental electron backscatter diffraction data of a fatigued metal film,where the aim is to develop a model, which consistently and physically meaningfullycaptures the metal’s microstructural changes that are caused by external loading.
M1b.3
MON18:15
I18:45
W211/12
Block coordinate descent method for ill-posed problemsLukas Neumann? (Institut für Grundlagen der techn. Wissenschaften, Universität Innsbruck)Simon Rabanser (Institut für Mathematik, Universität Innsbruck)Markus Haltmeier (Institut für Mathematik, Universität Innsbruck)
Block coordinate descent methods are iterative schemes characterized by perform-ing updates in the direction of subgroups of coordinates. Such methods proofed effi-cient in optimization problems and convergence in the objective function is known.However, convergence in the pre-image space, as typically needed in inverse prob-lems, is not known and does not hold in general. We present a convergence analysisfor linear ill-posed problems. Moreover we discuss numerical tests for a linear prob-lem as well as a nonlinear example from one step inversion in multi-spectral X-raytomography.
This talk is based on [1].
[1] S. Rabanser, L. Neumann, and M. Haltmeier, 2019. Analysis of the Block Co-ordinate Descent Method for Linear Ill-Posed Problems, Preprint available athttps://arxiv.org/abs/1902.04794.
54 M1: Industrial and Applied Mathematics
M1c.1
TUE10:00
I10:30
W211/12
Process Model for the Industrial Plate Leveling Operation with Special Empha-sis on Online ControlMichael Liebrecht? (RICAM, MathConsult GmbH)Ann-Kristin Baum (RICAM)Roman Wahl (uni software plus GmbH)Michael Aichinger (uni software plus GmbH)Erik Parteder (voestalpine Grobblech GmbH)
To produce high quality steel plates that meet the customers’ requirements, physi-cally based models are increasingly used in the automation systems of the productionprocess. In this work, a 2D model to describe the leveling process of the heavy plateproduction is presented. The model is based on Euler beam theory and featuresisotropic hardening of the material and the elastic response of the leveling machine.It enables the computation of the full stress-strain history as well as all contact pointsand contact forces between the plate and the leveling rolls. The results are validatedusing the multi-purpose finite ele-ment software ABAQUS and compared to mea-sured forces recorded during production. Changes of the mechanical properties ofthe plate caused by the leveling operation can be calculated from the stress-strainhistory. The implementation of the presented 2D model in C++ is dramatically fastercompared to the multi-purpose finite element solution. Nevertheless, due to timeconstraints during the production process, it is not feasible to apply the 2D model inreal time. For this purpose, a surrogate model based on vast pa-rameter studies hasbeen developed.
M1c.2
TUE10:30
I11:00
W211/12
Nonlinear inverse heat transfer problems in modelling the cooling process ofheavy platesDimitri Rothermel? (Universität des Saarlandes)Thomas Schuster (Universität des Saarlandes)
The goal of our project is the automation of the production of heavy plates fromsteel and the interest in adjusting the mechanical properties of those. For that reason,we want to model and control the cooling process depending on the water load andthe speed of the cooling plate. Based on this, we get an optimal control problem witha nonlinear PDE (heat transfer equation) constraint, describing the heat (or tempera-ture) distribution in the given region over time. This PDE however, is equipped withunknown functions which have to be characterized in different subproblems. In thistalk we will give an overview of the project, present the experimental setup and showsome numerical results.
M1: Industrial and Applied Mathematics 55
M1c.3
TUE11:00
I11:30
W211/12
Graph Based Methods for Hyperspectral Image ClassificationPhilip Sellars? (Department of Mathematics, University of Cambridge)Angelica I. Aviles Rivero (Department of Mathematics, University of Cambridge)Carola Bibiane Schönlieb (Department of Mathematics, University of Cambridge)
A central problem in hyperspectral image classification is obtaining high clas-sification accuracy when using a limited amount of labelled data. In this talk Iwill present a novel graph-based semi-supervised framework to tackle this problem,highlighting the power of graph based learning. Our framework uses a superpixel ap-proach, allowing it to define meaningful local regions in hyperspectral images, whichwith high probability share the same classification label. We then extract spectral andspatial features from these regions and use them to produce a contracted weightedgraph-representation, where each node represents a region rather than a pixel. Thegraph is then fed into a graph-based semi-supervised classifier which gives the fi-nal classification. We show that using superpixels in a graph representation is aneffective tool for speeding up graphical classifiers applied to hyperspectral images.We demonstrate through exhaustive quantitative and qualitative results that our pro-posed method produces accurate classifications when an incredibly small amount oflabelled data is used.
M1c.4
TUE11:30
I12:00
W211/12
Regularizing networks for soving inverse problemsJohannes Schwab? (Universität Innsbruck)Markus Haltmeier (Universität Innsbruck)Stephan Antholzer (Universität Innsbruck)
Recently, data driven methods for solving inverse problems have become verypopular. With the help of deep neural networks, these methods empirically showoutstanding performance but still lack of theoretical justification. In our work weanalyze such trained neural networks combined with a classical regularization as areconstruction method. We show, that for a particular type of networks, called regu-larizing networks, the proposed approach leads a convergent regularization method.Additionally we show convergence rates and illustrate the proposed method by theapplication of regularizing networks to limited view photoacoustic tomography.
56 M1: Industrial and Applied Mathematics
M1c.5
TUE12:00
I12:30
W211/12
Known Operator Learning — An Approach to unite Physics, Signal Processing,and Machine LearningAndreas Maier (Friedrich-Alexander-Universität Erlangen-Nürnberg)
We describe an approach for incorporating prior knowledge into machine learningalgorithms. We aim at applications in physics and signal processing in which weknow that certain operations must be embedded into the algorithm. Any operationthat allows computation of a gradient or sub-gradient towards its inputs is suited forour framework. We prove that the inclusion of such prior knowledge reduces max-imal error bounds. Furthermore, we demonstrate experimentally that it reduces thenumber of free parameters. We apply this approach to various tasks ranging fromCT image reconstruction over vessel segmentation to the derivation of previouslyunknown imaging algorithms. As such the concept is widely applicable for manyresearchers in physics, imaging, and signal processing. We assume that our analy-sis will support further investigation of known operators in other fields of physics,imaging, and signal processing.
[1] A. Maier, F. Schebesch, C. Syben, T. Würfl, S. Steidl, J.-H. Choi, and R. Fahrig, “Pre-cision Learning: Towards Use of Known Operators in Neural Networks,” in 2018 24rdInternational Conference on Pattern Recognition (ICPR), J. K. T. Tan, Ed., 2018, pp.183–188. [Online].
[2] C. Syben, B. Stimpel, J. Lommen, T. Würfl, A. Dörfler, and A. Maier, “Deriving neu-ral network architectures using precision learning: Parallel-to-fan beam conversion,” inGerman Conference on Pattern Recognition (GCPR), 2018.
Minisymposium M2Mathematical Finance in the age of MachineLearning
Josef Teichmann (ETH Zürich)
58 M2: Mathematical Finance in the age of Machine Learning
M2a.1
THU10:00
I10:30W205
Dynamic and Control Theoretical Aspects of Reservoir ComputingJuan-Pablo Ortega? (Universität Sankt Gallen, Switzerland and CNRS, France)Lukas Gonon (Universität Sankt Gallen, Switzerland)Lyudmila Grigoryeva (Universität Konstanz, Germany)
Reservoir computing (RC) carries out the learning of input/output systems by us-ing families of state space systems in which the training of a small portion of their pa-rameters suffices for excellent generalization properties. In this talk we shall analyzein detail the connection between various dynamic and control theoretical features ofwidely used families of RC families and their impact in basic learning theoreticalgoals, in particular in the development of bounds for the approximation and estima-tion errors and in the design of dimension reduction techniques.
[1] Lukas Gonon and Juan-Pablo Ortega: Reservoir computing universality with stochasticinputs. To appear in IEEE Transactions on Neural Networks and Learning Systems,2018.
[2] Lyudmila Grigoryeva and Juan-Pablo Ortega. Universal discrete-time reservoir comput-ers with stochastic inputs and linear readouts using non-homogeneous state-affine sys-tems. Journal of Machine Learning Research, 19 (24):1–40, 2018.
[3] Lyudmila Grigoryeva and Juan-Pablo Ortega. Echo state networks are universal. NeuralNetworks, 108: 495–508, 2018.
[4] Lyudmila Grigoryeva and Juan-Pablo Ortega. Differentiable reservoir computing.Preprint, 2019.
M2: Mathematical Finance in the age of Machine Learning 59
M2a.2
THU10:30
I11:00W205
Forecasting of Realized (Co)Variances with Reservoir ComputingLyudmila Grigoryeva? (Universität Konstanz, Germany)Lukas Gonon (Universität Sankt Gallen, Switzerland)Oleksandra Kukharenko (Universität Konstanz, Germany)Juan-Pablo Ortega (Universität Sankt Gallen, Switzerland and CNRS, France)
The problem of forecasting realized (co)variances (RV) computed out of intradayreturns of the components of the S&P 500 market index is considered. The studyfocuses on a novel machine learning paradigm known as reservoir computing (RC)for producing multistep ahead forecasts for time series of realized (co)variances.Various families of reservoir computers have been recently proved to have univer-sal approximation properties when processing stochastic discrete-time semi-infiniteinputs [1, 2, 3]. We examine the empirical performance of RC in forecasting of real-ized (co)variances in comparison to many conventional state-of-the-art econometricmodels for various RV estimators, periods, and dimensions. We show that universalRC families consistently demonstrate superior predictive ability for various designsof empirical exercises.
[1] Lyudmila Grigoryeva and Juan-Pablo Ortega. Universal discrete-time reservoir comput-ers with stochastic inputs and linear readouts using non-homogeneous state-affine sys-tems. Journal of Machine Learning Research, 19 (24):1–40, 2018.
[2] Lyudmila Grigoryeva and Juan-Pablo Ortega. Echo state networks are universal. NeuralNetworks, 108: 495–508, 2018.
[3] Lukas Gonon and Juan-Pablo Ortega: Reservoir computing universality with stochasticinputs. To appear in IEEE Transactions on Neural Networks and Learning Systems,2018.
M2a.3
THU11:00
I11:30W205
Error Bounds for Random Recurrent Neural Networks and General ReservoirComputing SystemsLukas Gonon? (Universität St. Gallen)Lyudmila Grigoryeva (Universität Konstanz)Juan-Pablo Ortega (Universität St. Gallen and CNRS)
In this talk we present our recent mathematical results on universality and errorbounds for reservoir computing. Motivated by the task of realized volatility fore-casting, we study approximation and learning based on random neural networks andmore general reservoir computing systems. For echo state networks with output feed-back we obtain high-probability bounds on the approximation error in terms of thenetwork parameters. For a more general class of reservoir computing systems andweakly dependent (possibly non-i.i.d.) input data, we then also derive generalizationerror bounds based on a Rademacher-type complexity.
M2a.4
THU11:30
I12:00W205
Missing abstract file directories/Teichmann-Josef/Teichmann-Josef-235.pdf !!!
60 M2: Mathematical Finance in the age of Machine Learning
M2a.5
THU12:00
I12:30W205
Universal Approximation TheoremsAnastasis Kratsios (ETH Zürich)
The universal approximation theorem established the density of specific familiesof neural networks in the space of continuous functions and in various Bochner-Lebesgue spaces, defined between any two Euclidean spaces. We extend and refinetheir result and definitions by proving that there exist dense neural network architec-tures on a much larger class of function spaces and that these architectures may bewritten down using only a finite number of functions. Our results cover any functionspace which is homeomorphic either to a convex body in an infinite-dimensionalFréchet space, to a finite-dimensional compact Riemannian manifold, or to a metricspace which is a retract of a Banach space. We prove that upon appropriately ran-domly selecting the neural networks architecture’s activation function we may stillobtain a dense set of neural networks, with positive probability.
Conversely, we show that given any neural network architecture on a set of con-tinuous functions between two T0 topological spaces, there exists a unique finesttopology on that set of functions which makes the neural network architecture into auniversal approximator. Several examples are considered throughout.
M2b.1
THU15:30
I16:00W205
Randomized shallow neural networks and their use in understanding gradientdescentHanna Wutte? (ETH Zürich)Jakob Heiss (TU Wien)Josef Teichmann (ETH Zürich)
Today, various forms of neural networks are trained to perform approximationtasks in many fields (including Mathematical Finance). However, the solutions ob-tained are not wholly understood. Empirical results suggest that the training favorsregularized solutions. Moreover, it has been questioned how much training reallymatters, in the sense that randomly choosing subsets of the network’s weights andtraining only a few leads to an almost equally good performance.
These observations motivate us to analyze properties of the solutions found by thegradient descent algorithm frequently employed to perform the training task. In par-ticular, we consider one dimensional (shallow) neural networks in which weights arechosen randomly and only the last layer is trained. We show, that the resulting solu-tion converges to the smooth spline interpolation of the training data as the numberof hidden nodes tends to infinity. This might give valuable insight on the propertiesof the solutions obtained using gradient descent methods in general settings.
M2: Mathematical Finance in the age of Machine Learning 61
M2b.2
THU16:00
I16:30W205
Modeling rough covariance processesChrista Cuchiero? (Wirtschaftsuniversität Wien)Josef Teichmann (ETH Zürich)
The rough volatility paradigm asserts that the trajectories of assets’ volatility arerougher than Brownian motion, a revolutionary perspective that has changed certainpersistent views of volatility. It provides via stochastic Volterra processes a universalapproach to capture important features of time series and option price data as wellas microstructural foundations of markets. We provide an infinite dimensional pointof view on stochastic Volterra processes, which allows to dissolve a generic non-Markovanity of the at first sight naturally low dimensional volatility process. Thisapproach enables to go beyond the univariate case considered so far and to treat thechallenging problem of multivariate rough covariance models, in particular of affineand Wishart type, for more than one asset.
62 M2: Mathematical Finance in the age of Machine Learning
M2b.3
THU16:30
I17:00W205
A Neural Network Approach to Local Stochastic Volatility CalibrationWahid Khosrawi? (ETH Zurich)Christa Cuchiero (Vienna University of Economics and Business)Josef Teichmann (ETH Zurich)
A central task in modeling, which has to be performed each day in banks andfinancial institutions, is to calibrate models to market and historical data. So farthe choice which models should be used was not only driven by their capacity ofcapturing empirically the observed market features well, but rather by computationaltractability considerations. Due to recent work in the context of machine learning,this notion of tractability has changed significantly. In this work, we show howa neural network approach can be applied to the calibration of (multivariate) localstochastic volatility (LSV) models. In these models (assuming for simplicity zerointerest rates), the price process (St)t≥0 satisfies
dSt = StL(t,St)αtdWt ,
where (αt)t≥0 is some stochastic process taking values in R. The function L is theso-called leverage function and the goal is to determine L such that option pricesobserved in the financial market are perfectly calibrated. Under strong assumptionson the observed prices - e.g. it is assumed that a continuum of prices are availablewhich requires interpolation of observations - it is possible to perfectly calibrate thecorresponding local volatility model by means of the Dupire-formula, see [1]. Inpractice, a McKean-Vlasov SDE type approach is then used to calibrate the corre-sponding SLV model, which by itself is a delicate task, compare [2]. We show how adirect calibration of the LSV model is possible by employing a neural network typeapproach without the need of interpolation methods for the financial data.
[1] B. Dupire, Pricing With a Smile, Risk 7, (1994) 18 - 20.[2] J. Guyon and P. Henry-Labordere, The smile calibration problem solved, Preprint ,
(2011).
M2: Mathematical Finance in the age of Machine Learning 63
M2c.1
THU17:30
I18:00W205
Deep ALMThomas Krabichler? (Hochschule Luzern)Josef Teichmann (ETH Zürich)
There are very promising proofs-of-concept that reinforcement learning may rev-olutionise valuations and risk management strategies in the financial industry. Acornerstone in this regard is the concept "deep hedging". Accounting for transactioncost and liquidity constraints in the hedging of a multi-dimensional derivative expo-sure cannot be tackled at all with conventional methods from mathematical finance.Utilising techniques inspired from re-inforcement learning allows to overcome boththe complexity and the curse of dimension. We aim at refining this approach forentire term structures in order to solve classical yet intractable problems from asset-liability-management (ALM).
M2c.2
THU18:00
I18:30W205
Infinite dimensional polynomial jump-diffusionsSara Svaluto-Ferro? (University of Vienna)Christa Cuchiero (University of Vienna)
We introduce polynomial jump-diffusions taking values in an arbitrary Banachspace via their infinitesimal generator. We obtain two representations of the (condi-tional) moments in terms of solution of systems of ODEs. These representations gen-eralize the well-known moment formulas for finite dimensional polynomial jump-diffusions. We illustrate the practical relevance of these formulas by several appli-cations. In particular, we consider (potentially rough) forward variance polynomialmodels and we illustrate how to use the moment formulas to compute prices of VIXoptions.
64 M2: Mathematical Finance in the age of Machine Learning
Minisymposium M3Parallel and Distributed Optimization andSimulation
Steffen Finck (Fachhochschule Vorarlberg)
66 M3: Parallel and Distributed Optimization and Simulation
M3a.1
MON15:15
I15:45U406
Predicting the efficiency of LRZ’s cooling infrastructure using machinelearningHayk Shoukourian (Leibniz Supercomputing Centre (LRZ) of the Bavarian Academyof Sciences and Humanities; Ludwig-Maximilians-Universität München (LMU))
The power consumption of modern High Performance Computing (HPC)and cloud data centers increases hand-in-hand with their performance. Thementioned power consumption does not only translate to high operationalcosts but also to high carbon footprints, affecting the environment and limit-ing the further expansion of data centers, leading to an introduction of certainenergy/power capping strategies. All these make it important to be preemp-tive in improving energy/power efficiency of HPC and cloud data centersthat encompass the complete chain of power/energy consumers ranging fromdata center’s building infrastructure over compute resource management andscheduling to application scaling.
The talk will give a brief overview on LRZ’s newly deployed SuperMUC-NG flagship supercomputing system, outline its energy-efficiency character-istics as well as present a machine learning based approach used for modelingthe efficiency of HPC data center’s chiller-less hot water cooling loop. Thetalk will also highlight the applications of the developed model for the im-provement of energy-efficiency in modern data centers.
Keywords: machine learning, energy-efficiency, cooling efficiency, data cen-ter, HPC
M3: Parallel and Distributed Optimization and Simulation 67
M3a.2
MON15:45
I16:15U406
Towards multi-objective, region-based auto-tuning for parallel programsPhilipp Gschwandtner (University of Innsbruck, Austria)
Efficient parallelization and optimization for modern parallel architectures is atime-consuming and error-prone task that requires numerous iterations of code trans-formations, tuning and performance analysis which in many cases have to be redonefor each different architecture. In this talk we introduce a novel auto-tuning com-piler named INSIEME which consists of a compiler component featuring a multi-objective optimizer and a runtime system. The multi-objective optimizer derives aset of non-dominated solutions, each of them expressing a trade-off among differentconflicting objectives such as runtime, resource efficiency and energy consumption.This set is commonly known as Pareto set. Our search algorithm, which explorescode transformations and their parameter settings, is based on differential evolution.Additionally, rough sets are employed to reduce the search space, and thus the num-ber of evaluations required during compilation. We further examine global versusregion-based tuning techniques and demonstrate our approach by tuning shared me-mory parallel programs for execution time, energy consumption and resource usage.
68 M3: Parallel and Distributed Optimization and Simulation
M3a.3
MON16:15
I16:45U406
Optimization with the Distributed Execution FrameworkThomas Feilhauer? (Fachhochschule Vorarlberg)Steffen Finck (Fachhochschule Vorarlberg)
A central task of the Distributed Execution Framework (DEF) [1] is to supplyefficient library routines across applications for simulation and optimization. Theroutines can then be included by the users in their applications and can be executedin parallel on the DEF. A key issue that is important for the acceptance of the DEFamong users and routine developers is the independence of the programming lan-guage and the computer platform. Developers can implement their routines in anyprogramming language, and the DEF users can also write their client applicationsin any programming language independent of the library routines used therein. Theroutines are called from the client application via an API developed for the DEF andthe DEF, which is connected to the client application via a web service interface,takes care of the processing of the library routine calls.
The other important feature of the DEF is support for parallel processing of DEFapplications. The DEF permits to create clusters with any number of worker nodes.For this purpose, the DEF relies on cloud technologies that enables the clusters tobe operated both On- and Off-Premise. Finally, at runtime, the DEF distributes theinvocation of library routines from the client application to the worker instancesprovided by the cluster.
The DEF’s capabilities are exemplified on an expensive optimization problem. Theaim is to perform a simulation-based optimization on a model of a bank’s balancesheet to determine the worst case scenario [2]. The problem contains various boundconstraints, a non-linear constraint, and is high-dimensional. Two approaches areused for the optimization, a simple random search approach and a Evolution Strategy.Next to the results obtained by these approaches, the respective run times and thescalability of the DEF are investigated.
[1] T. Feilhauer and M. Sobotka, DEF-a programming language agnostic framework andexecution environment for the parallel execution of library routines, Journal of CloudComputing, 5(1) (2016)
[2] S. Finck, Worst Case Search over a Set of Forecasting Scenarios Applied to FinancialStress-Testing, GECCO 19 Companion, ACM, (2019)
M3: Parallel and Distributed Optimization and Simulation 69
M3b.1
MON17:15
I17:45U406
Analyzing Autonomous Demand Side Algorithms with Parallel ComputationFrameworksKlaus Rheinberger? (Research Center Energy and Josef Ressel Centre for Applied ScientificComputing, FH Vorarlberg)Ramona Rosskopf (Research Center Energy and Josef Ressel Centre for Applied ScientificComputing, FH Vorarlberg)Markus Preißinger (Research Center Energy, FH Vorarlberg)Peter Kepplinger (Research Center Energy and Josef Ressel Centre for Applied ScientificComputing, FH Vorarlberg)
In contrast to fossil energy sources, the supply by renewable energy sources likewind and photovoltaics can not be controlled. Therefore, flexibilities on the demandside of the electric power grid, like electro-chemical energy storage systems, are usedincreasingly to match electric supply and demand at all times. To control those flex-ibilities, we consider two algorithms that both lead to linear programming problems.These are solved autonomously on the demand side, i.e., by household computers.
In the classic approach, an energy price signal is sent by the electric utility to thehouseholds, which, in turn, optimize the cost of consumption within their constraints.Instead of an energy price signal, we claim that an appropriate power signal that istracked in L1-norm as close as possible by the household has favorable character-istics. We argue that an interior point of the household’s feasibility region is neveran optimal price-based point but can result in a L1-norm optimal point. Thus, pricesignals can not parametrize the complete feasibility region which may not lead to anoptimal allocation of consumption.
We compare the price and power tracking algorithms over a year on the base ofone-day optimizations regarding different information settings and using a large dataset of daily household load profiles. The computational task constitutes an embar-rassingly parallel problem. To this end, the performance of the two parallel compu-tation frameworks DEF [1] and Ray [2] are investigated. The Ray framework is usedto run the Python applications locally on several cores. With the DEF frameworkwe execute our Python routines parallelly in a cloud. All in all, the results providean understanding of when which computation framework and autonomous algorithmwill outperform the other.
[1] T. Feilhauer, M. Sobotka, DEF-a programming language agnostic framework and exe-cution environment for the parallel execution of library routines, Journal of Cloud Com-puting (2016) 5: 20. https://doi.org/10.1186/s13677-016-0070-z
[2] P, Moritz, R. Nishihara et al., Ray: A Distributed Framework for Emerging AI Applica-tions, arXiv:1712.05889v2 [cs.DC] (2017). https://arxiv.org/abs/1712.05889
70 M3: Parallel and Distributed Optimization and Simulation
Minisymposium M4Spectral Theory
Gerald Teschl (Universität Wien)Jussi Behrndt (Universität Graz)
72 M4: Spectral Theory
M4a.1
THU10:00
I10:30W206
On the integrability of the Benjamin-Ono equation with periodic boundary con-ditionsThomas Kappeler (University of Zurich)
In this talk I report on joint work with Patrick Gérard (Université Paris-Sud) con-cerning the construction of global Birkhoff coordinates for the Benjamin-Ono equa-tion. In these coordinates this equation can be solved by quadrature, meaning thatit is an integrable (pseudo)differential equation in the strongest possible sense. Theconstruction is based on the Lax pair formulation of this equation. I will present spec-tral properties of the Lax operator, discuss a generating function of the Benjamin-Ono hierarchy, which allows to establish various trace formulas, and introduce theBirkhoff coordinates. Furthermore, I will provide a characterization of finite gapsolutions.
M4a.2
THU10:30
I11:00W206
Uniformity of cocycles and spectra of Schreier graphsDaniel Lenz (Friedrich-Schiller University Jena)
We introduce a class of subshifts governed by finitely many two-sided infinitewords. We show that any locally constant cocycle over such a subshift is uniform.From this we obtain Cantor spectrum of Lebesgue measure zero for the associatedJacobi operators if the subshift is aperiodic. Our class covers all simple Toeplitzsubshifts as well as all Sturmian subshifts. We apply our results to the spectral theoryof Schreier graphs for uncountable families of groups acting on rooted trees. (Jointwork with Rostislav Grigorchuk, Tatiana Nagnibeda, Daniel Sell).
M4a.3
THU11:00
I11:30W206
On the absolutely continuous spectrum of generalized indefinite stringsAleksey Kostenko (University of Ljubljana and University of Vienna)
Our main result is the stability of the absolutely continuous spectrum of two modelexamples of (Krein) strings under rather wide classes of perturbations. In particu-lar, this enables us to prove that the absolutely continuous spectrum of the isospec-tral problem associated with the conservative Camassa–Holm flow in the dispersiveregime is essentially supported on the interval [1/4,∞) for the entire natural phasespace. If time permits, applications to quantum graphs and 1D Hamiltonians withδ ′-interactions will be discussed as well.
The talk is based on joint work with J. Eckhardt (Loughborough, UK).
M4: Spectral Theory 73
M4a.4
THU11:30
I12:00W206
Strichartz estimates for the one-dimensional wave equationRoland Donninger? (Universität Wien)Irfan Glogic (Universität Wien)
We consider the hyperboloidal initial value problem for the one-dimensional waveequation, perturbed by a smooth potential. We show that the evolution decomposesinto a finite-dimensional part that is controlled by spectral theory and an infinite-dimensional part that satisfies Strichartz estimates, provided a certain spectral as-sumption holds. As an application we consider the long-time asymptotics of Yang-Mills fields on a wormhole spacetime.
M4a.5
THU12:00
I12:30W206
Discrete Dirac systems: spectral and scattering theories and Verblunsky-typecoefficientsAlexander Sakhnovich (University of Vienna)
We consider discrete Dirac systems, give explicit expressions of rational Weylfunctions and reflection coefficients, solve explicitly inverse problems and study sta-bility of the procedure. We introduce also new Verblunsky-type coefficients andobtain Verblunsky-type results for Dirac systems and corresponding block Toeplitzmatrices. The results were published in
[1] B. Fritzsche, B. Kirstein, I. Roitberg, A.L. Sakhnovich, Discrete Dirac systems on thesemiaxis: rational reflection coefficients and Weyl functions, J. Difference Equ. Appl. 25(2019) 294–304.
[2] I. Roitberg, A.L. Sakhnovich, The discrete self-adjoint Dirac systems of general type:explicit solutions of direct and inverse problems, asymptotics of Verblunsky-type coeffi-cients and the stability of solving of the inverse problem, Zh. Mat. Fiz. Anal. Geom. 14(2018) 532–548.
[3] A.L. Sakhnovich, New "Verblunsky-type” coefficients of block Toeplitz and Hankel ma-trices and of corresponding Dirac and canonical systems, J. Approx. Theory 237 (2019)186–209.
74 M4: Spectral Theory
M4b.1
THU15:30
I16:00W206
Threshold Singularities of the Spectral Shift Function for Geometric Perturba-tions of a Magnetic HamiltonianGeorgi Raikov (Pontificia Universidad Católica de Chile)
I will consider the 3D Schrödinger operator H0 with constant magnetic field, andits perturbations H+ (resp., H− ) obtained from H0 by imposing Dirichlet (resp.,Neumann) conditions on an appropriate surface. I will introduce the Krein spectralshift function for the operator pairs (H+,H0) and (H−,H0), and will discuss its sin-gularities at the Landau levels which play the role of thresholds in the spectrum ofthe unperturbed operator H0.The talk is based on a joint work with Vincent Bruneau (Bordeaux).The financial support of the Chilean science fund Fondecyt under grant 1170816 isgratefully acknowledged.
M4b.2
THU16:00
I16:30W206
Rarefaction and shock waves for the Toda equationJohanna Michor? (Universitat Wien)Iryna Egorova (ILT Kharkiv)Gerald Teschl (Universität Wien)
We give a survey of the long-time asymptotic of solutions of the Toda lattice equa-tion with steplike constant initial data using the nonlinear steepest descent analysisfor oscillatory Riemann–Hilbert factorization problems and its extension based on asuitably chosen g-function. Analytic formulas for the leading term of the asymptoticsolutions of the Toda shock and rarefaction problems are given and complemented bynumerical simulations. We provide an explicit formula for the modulated solution interms of Abelian integrals on the underlying hyperelliptic Riemann surface and thusverify a conjecture from Venakides, Deift, and Oba from 1991. The correspondingresults are published in [1, 2, 3].
[1] J. Michor, Wave phenomena of the Toda lattice with steplike initial data, Phys. Lett. A380 (2016) 1110–1116.
[2] I. Egorova, J. Michor, and G. Teschl, Rarefaction waves for the Toda equation via non-linear steepest descent, Discrete Contin. Dyn. Syst. 38(4) (2018) 2007–2028.
[3] I. Egorova, J. Michor, and G. Teschl, Long-time asymptotics for the Toda shock problem:non-overlapping spectra, Zh. Mat. Fiz. Anal. Geom. 14(4) (2018) 406–451.
M4: Spectral Theory 75
M4b.3
THU16:30
I17:00W206
Riemann–Hilbert problems in asymptotic analysisMateusz Piorkowski? (University of Vienna)Iryna Egorova (B. Verkin ILTPE, Kharkiv)Gerald Teschl (University of Vienna)
Our research revolves around the applications of the Riemann–Hilbert (R-H) for-malism to problems from asymptotic analysis. In particular, we have been analysingthe long time behaviour of solutions to the KdV equation with step like initial data.We observed a previously unknown singular behaviour of the R-H solutions, whichposes a new challenge to the usual approach. Our solution is based on a more carefulanalysis of the connection between R-H problems and singular integral equations.Similar ideas can be used in the long time asymptotics of orthogonal polynomials ona finite interval.
[1] P. Deift and X. Zhou, A steepest descent method for oscillatory Riemann–Hilbert prob-lems, Ann. of Math. (2) 137, 295–368 (1993).
[2] I. Egorova, Z. Gladka, V. Kotlyarov and G. Teschl, Long-time asymptotics for theKorteweg–de Vries equation with step-like initial data, Nonlinearity 26 (2013), no. 7,1839-1864.
M4c.1
THU17:30
I18:00W206
Product formulas and convolutions for solutions of Sturm-Liouville equationsRúben Sousa? (CMUP, Faculty of Sciences, University of Porto)Manuel Guerra (CEMAPRE and ISEG, University of Lisbon)Semyon Yakubovich (CMUP, Faculty of Sciences, University of Porto)
The Fourier transform, which lies at the heart of the classical theory of harmonicanalysis, is generated by the eigenfunction expansion of the Sturm-Liouville operator− d2
dx2 . This naturally raises a question: is it possible to generalize the main facts ofharmonic analysis to integral transforms of Sturm-Liouville type?
In this talk we introduce a novel unified framework for the construction of productformulas and convolutions associated with a general class of regular and singularSturm-Liouville boundary value problems. This unified approach is based on the ap-plication of the Sturm-Liouville spectral theory to the study of the associated hyper-bolic equation. As a by-product, an existence and uniqueness theorem for degeneratehyperbolic Cauchy problems with initial data at a parabolic line is established.
We will show that each Sturm-Liouville convolution gives rise to a Banach algebrastructure in the space of finite Borel measures in which various probabilistic conceptsand properties can be developed in analogy with the classical theory. Exampleswill be given, showing that many known convolution-type operators — such as theHankel, Jacobi and Whittaker convolutions — can be constructed using this generalapproach.
76 M4: Spectral Theory
M4c.2
THU18:00
I18:30W206
Optimal Blowup Stability for the Energy Critical Wave EquationZiping Rao? (Universität Wien)Roland Donninger (Universität Wien)
We study the stability of ODE blowup solutions of wave equations with powernonlinearity in the lightcone. Perturbing around the ODE blowup solution, we obtaina nonlinear wave equation with a time-dependent self-similar potential. In order tohandle this potential, we introduce similarity coordinates, under which we rewritethe equation into an autonomous evolution equation. We first recall some stabilityresults in higher regularity by Donninger and Schörkhuber. They use the Lumer–Phillips theorem to obtain a solution semigroup to the Cauchy problem. Then by theGearhart–Prüss theorem they obtain enough decay of the semigroup to control thenonlinearity.
For the energy critical equation in the energy space, the semigroup does not haveenough decay to control the nonlinearity. For this we need the help of Strichartzestimates. By constructing an explicit expression of the semigroup, we establishthese estimates, and moreover improve the energy bound from the Gearhart–Prüsstheorem. Following the pioneering work of the three dimensional case by Donningerin [?], we are able to obtain energy critical blowup stability in five dimensions in [?].
[1] R. Donninger, Strichartz estimates in similarity coordinates and stable blowup for thecritical wave equation, Duke Math. J., Volume 166, Number 9 (2017), 1627-1683.
[2] R. Donninger, Z. Rao Blowup stability at optimal regularity for the critical wave equa-tion, arXiv:1811.08130, 2018.
Minisymposium M5Inverse Problems for Partial DifferentialEquations
William Rundell (Texas A&M University)Barbara Kaltenbacher (Universität Klagenfurt)
78 M5: Inverse Problems for Partial Differential Equations
M5a.1
WED11:00
I11:30
W211/12
NETT: Solving Inverse Problems with Deep Neural NetworksMarkus Haltmeier? (Department of Mathematics, University Innsbruck)Johannes Schwab (Department of Mathematics, University Innsbruck)Stephan Antholzer (Department of Mathematics, University Innsbruck)Housen Li (Institut für Mathematische Stochastik, Universität Göttingen)
Recently, deep learning and neural network based algorithms appeared as newparadigm for solving inverse problems. In this talk we propose and analyze vari-ational methods for inverse problems using a neural network as regularizer. Wepresent a convergence analysis, derive convergence rates, and propose a possibletraining strategy. Numerical results are presented where the network approach demon-strates good performance even for unknowns very different from the training data.
This talk is based on [1].
[1] H. Li, J. Schwab, S. Antholzer, and M. Haltmeier, M., 2018. NETT: Solving inverseproblems with deep neural networks. arXiv:1803.00092.
M5a.2
WED11:30
I12:00
W211/12
Sparse synthesis regularization using deep neural networksDaniel Obmann? (University of Innsbruck)Markus Haltmeier (University of Innsbruck)Johannes Schwab (University of Innsbruck)
We present a sparse synthesis regularization method for solving inverse problems.To achieve sparsity we use an encoder-decoder network which was trained with anadded `1 regularization term. We demonstrate that images similar to the trainingdata can be recovered from their thresholded synthesis coefficients using the traineddecoder.To solve inverse problems using this method we propose minimizing an `1-Tikhonovfunctional which is a sum of a data-fitting term and a weighted `1 regularization termwhich regularizes the synthesis coefficients.
M5a.3
WED12:00
I12:30
W211/12
Singular Value Decomposition for Atmospheric TomographyRonny Ramlau? (Kepler University Linz and RICAM)Simon Hubmer (RICAM)Andreas Neubauer (Kepler University Linz)
In atmospheric tomography, light originating from guide stars travels through theatmosphere. Its recorded wavefronts are used to reconstruct the turbulence in theatmosphere. The information on the turbulence is then used to obtain sharp imagesfrom astronomical telescopes. Taking into account the layered structure of the turbu-lent atmosphere, the properties of the guide stars as well as the geometric structureof the telescope, we present singular value decompositions of the underlying tomog-raphy operators for different telescope settings.
M5: Inverse Problems for Partial Differential Equations 79
M5b.1
THU10:00
I10:30
W211/12
Determining nonlinear terms in a reaction-diffusion systemWilliam Rundell (Texas A&M University)
Reaction-diffusion equations are one of the most common partial differential equa-tions used to model physical phenomenon. They arise as the combination of twophysical processes: a driving force f (u) that depends on the state variable u and adiffusive mechanism that spreads this effect over a spatial domain. The canonicalform is ut −4u = f (u). Application areas include chemical processes, heat flowmodels and population dynamics. As part of the model building, assumptions aremade about the form of f (u) and these inevitably contain various physical constants.The direct or forwards problem for such equations is now very well-developed andunderstood, especially when the diffusive mechanism is governed by Brownian mo-tion resulting in an equation of parabolic type.
However, our interest lies in the inverse problem of recovering the reaction termf (u) not just at the level of determining a few parameters in a known functionalform, but recovering the complete functional form itself. To achieve this we set upthe usual paradigm for the parabolic equation where u is subject to both given initialand boundary data, then prescribe overposed data consisting of the solution at a latertime T . We will transform the inverse problem into an equivalent nonlinear mappingfrom which we seek a fixed point. We will be able to prove important featuresof this map such as a self-mapping property and give conditions under which it iscontractive.
Finally, we will indicate modifications required to incorporate more complex dif-fusion processes than that offered by classical Brownian motion.
80 M5: Inverse Problems for Partial Differential Equations
M5b.2
THU10:30
I11:00
W211/12
Identification and Design of EM-Chiral Scattering ObstaclesFrank Hettlich (Karlsruhe Institute of Technology (KIT))
In scattering theory the far field operator includes to some extent geometrical andphysical properties of the scattering object. Thus, inverse problems like recoveringthe shape of an obstacle from a far field pattern are of particular interest. As a firststep a linearization of the far field pattern with respect to the shape of a penetrableobstacle in case of electromagnetic waves is discussed. Characterizing the corre-sponding domain derivative gives access to the linearization of the far field operator.Solving corresponding boundary integral equations with a sophisticated applicationof the boundary element platform Bempp illustrates the feasibility of regularizedNewton type schemes in solving such highly ill-posed problems.
Furthermore, a geometry depending property of a scatterer is its chirality. Thenotion of electromagnetic chirality, recently introduced in the Physics literature, usesthe correspondence of scatterer and far field pattern to define a quantifiable measureof chirality. We show that this measure can be given by the Hilbert-Schmidt normsof certain projections of the far field operator. An investigation of this splitting of theoperator together with the above linearization with respect to geometry leads to thepossibility in identifying and designing optimal chiral scattering obstacles, which areof special importance in chemistry and in physics. A first attempt to the challengingoptimization problem in case of a fixed wave number is introduced and discussed.
M5b.3
THU11:00
I11:30
W211/12
Uniqueness and reconstruction in a passive inverse problem of helioseismologyAlexey Agaltsov (Max-Planck-Institut für Sonnensystemforschung, Göttingen)
We consider the inverse problem of recovering the radially symmetric sound speed,density and attenuation in the Sun from the measurements of the solar acoustic fieldat two heights above the photosphere and for a finite number of frequencies abovethe acoustic cutoff frequency. We show that this problem reduces to recovering along range potential (with a Coulomb-type decay at infinity) in a Schrödinger equa-tion from the measurements of the imaginary part of the radiation Green’s functionat two distances from zero. We demonstrate that generically this inverse problemfor the Schrödinger equation admits a unique solution, and that the original inverseproblem for the Sun admits a unique solution when measurements are performedat at least two different frequencies above the cutoff frequency. These uniquenessresults are confirmed by numerical simulations.
This talk is based on a joint work in progress with T. Hohage and R. G. Novikov.
M5: Inverse Problems for Partial Differential Equations 81
M5b.4
THU11:30
I12:00
W211/12
Parameter identification for the Landau-Lifshitz-Gilbert equation in MagneticParticle ImagingAnne Wald? (Saarland University, Saarbrücken)Tram Thi Ngoc Nguyen (Alpen-Adria-Universität Klagenfurt)Barbara Kaltenbacher (Alpen-Adria-Universität Klagenfurt)Thomas Schuster (Saarland University, Saarbrücken)
Magnetic particle imaging (MPI) is a relatively new tracer-based imaging tech-nique suited for medical imaging of, e.g., blood vessels [1]. Magnetic nanoparticlesare injected into the blood stream and serve as a tracer. In order to locate the parti-cles, a time-varying external magnetic field with a field-free point (FFP) is applied tothe object. More precisely, the FFP is moved through the object, causing a nonlinearresponse from the particles that are situated at the current position of the FFP. Thisresponse is measured as an induced voltage in the scanner’s receive coils. The goal isto reconstruct the particle concentration from these measurements. Mathematically,the forward problem is formulated as an integral equation of the first kind, where theintegration kernel is called the system function.
Prior to the actual imaging process, the scanner has to be calibrated, i.e., the sys-tem function has to be determined. This is currently done by a time-consumingpixel-wise scan of a known delta sample. Our goal now is a model-based approachfor the calibration that incorporates the relaxation behaviour of the nanoparticles.The Landau-Lifshitz-Gilbert equation is used as a physical model to describe themagnetization of the particles [2], which then yields the system function. We thusconsider a parameter identification problem for the Landau-Lifshitz-Gilbert equa-tion, were the measured currents for known samples are used as data [3]. The prob-lem is formulated and analyzed both in an all-at-once as well as in a reduced setting,see also [4, 5].
[1] T. Knopp, T. Buzug Magnetic Particle Imaging: an Introduction to Imaging Principlesand Scanner Instrumentation, Springer Berlin Heidelberg, 2012.
[2] M. Kruzik, A. Prohl Recent Developments in the Modeling, Analysis, and Numerics ofFerromagnetism, SIAM Review, 48(3) (2006) 439–483.
[3] B. Kaltenbacher, T. T. N. Nguyen, T. Schuster, A. Wald Parameter identification for theLandau-Lifshitz-Gilbert equation in Magnetic Particle Imaging, in preparation.
[4] B. Kaltenbacher, All-at-once versus reduced iterative methods for time dependent in-verse problems, Inverse Problems, 33(6) (2017) 064002.
[5] T. T. N. Nguyen, Landweber–Kaczmarz for parameter identification in time-dependentinverse problems: all-at-once versus reduced version, Inverse Problems, 35(3) (2019)035009.
82 M5: Inverse Problems for Partial Differential Equations
M5b.5
THU12:00
I12:30
W211/12
On convergence of Total Variation regularized inverse problemsGwenael Mercier (Universität Wien)
We extend some recent results on the Hausdorff convergence of level-sets for to-tal variation regularized linear inverse problems, as the noise and the regularizationparameter simultaneously vanish. Such a mode of convergence, although seldomused, is of particular interest in the context of recovery of piecewise constant coeffi-cients as well as in the processing of images composed mainly of objects separatedby clear boundaries. In these situations, Hausdorff convergence of level-sets can beseen as uniform convergence of the geometrical objects appearing in the data. Wewill consider dimensions higher than two and measurements in Banach spaces andinvestigate the relation between the dimension and the assumed integrability of thesolution that makes such a result hold. We also give some counterexamples of prac-tical application scenarios where the natural choice of fidelity term makes such aconvergence fail. Finally, we will shortly discuss the source condition assumptionunder which our result holds. This is a collaboration with J.A. Iglesias (RICAM,Linz).
M5: Inverse Problems for Partial Differential Equations 83
M5c.1
THU15:30
I16:00
W211/12
Generalized tolerance regularizationIwona Piotrowska-Kurczewski? (Universität Bremen)Georgia Sfakianaki (Universität Bremen)Peter Maaß (Universität Bremen)
The Tikhonov regularization of inverse problems is based on minimizing function-als Φ(x) = dist(F(x),yδ )+αR(x), which balance a discrepancy term dist(F(x),yδ ).It is obtained by measuring the mismatch between the data side and a regularizationor penalty term R(x), which encodes a-priori knowledge about the solution. TakingL2 norm of both terms leads to the classical Tikhonov regularization case which isvery well studied. A special case with sparsity constraints (L1 norm) is also wellestablished by now, see [1].
Motivated by applications in engineering we adjust the discrepancy and penaltyterms by including a tolerance measure, so called ε distance function, which ne-glects deviations from the data within a prescribed ε tolerance. This way we takeadvantage of the experts knowledge about y and or x. In particular, for linear oper-ators the transiton ε → 0 and p,q → 1 is equivalent to the transition from Tikhonovregularization to concept of support vector regression (SVR). This approach is ofimportance for applications where multiple measurements of the same object areavailable. It provides a confidence area for the true data. The parameter identifica-tion of (partial) differential equation is another application field for this technique.Additionally a smoothing property can be obtained for Tikhonov regularization con-taining the tolerances to the Lp-discrepancy term contains tolerances.
We present the analytical properties of xδα,ε for the Tikhonov regularization with
tolerances in either the discrepancy distε(F(x),yδ ) or the penalty term Rε(x). Weprove that in the limit point ε = 0 the classical Tikhonov minimizer is obtained.Numerical examples show a potential of this approach.
[1] M. Grasmair, M. Haltmeier, and O. Scherzer, Sparse regularization with lq penalty term.Inverse Problems, 24:055020 (13pp), 2008
84 M5: Inverse Problems for Partial Differential Equations
M5c.2
THU16:00
I16:30
W211/12
Dynamical super-resolution with applications to Ultrafast ultrasound imagingFrancisco Romero Hinrichsen? (University of Graz)Habib Ammari (ETH Zürich)Giovanni Alberti (University of Genoa)Timothée Wintz (ENS Paris)
There was a successful development in ultrasound imaging [1], that increased sig-nificantly its sampling rate and enhanced this imaging’s capacities. In particular, forvessel imaging, the use of microbubble tracking allows us to super-resolve blood ves-sels, and by estimating the particles’ speeds inside them, it is possible to calculate thevessels’ diameters [2]. In this context, following the theory of super-resolution [3],we model the microbubble tracking problem, formulating it in terms of a sparse spikerecovery problem in the phase space (the position and velocity space), that allows usto obtain simultaneously the speed of the microbubbles and their location. This leadsto an L1 minimization algorithm for point source tracking, that promises to be fasterthan current alternatives.
[1] M. Tanter and M. Fink, Ultrafast imaging in biomedical ultrasound, IEEE transactionson ultrasonics, ferroelectrics, and frequency control, 2014.
[2] C. Errico et Al., Ultrafast ultrasound localization microscopy for deep super-resolutionvascular imaging, Nature, 2015.
[3] E. Candès and C. Fernandez-Granda, Towards a Mathematical theory of Super-resolution, Communications on Pure and Applied Mathematics, 2014.
M5c.3
THU16:30
I17:00
W211/12
Regularizing sequential subspace optimization for the identification of the storedenergy of a hyperelastic structureRebecca Klein? (Saarland University)Anne Wald (Saarland University)Thomas Schuster (Saarland University)
We consider the nonlinear dynamic inverse problem of identifying the stored en-ergy function of hyperelastic materials from surface sensor measurements. In con-nection with the detection of damages in structures consisting of such materials thistask is highly interesting since all mechanical properties are hidden in the stored en-ergy function. This problem has already been solved using the attenuated Landwebermethod. Since this process is extremely time-consuming, we use sequential subspaceoptimiziation as a regularization technique.
M5: Inverse Problems for Partial Differential Equations 85
M5d.1
THU17:30
I18:00
W211/12
Solving the inverse problem of linear elasticity with monotonicity methodsSarah Eberle? (Goethe-University Frankfurt, Germany)Bastian Harrach (Goethe-University Frankfurt, Germany)Houcine Meftahi (ENIT of Tunisia, Tunisia)Taher Rezgui (ENIT of Tunisia, Tunisia)
The main motivation of this problem is the non-destructive testing of elastic struc-tures for material inclusions. We consider the inverse problem of recovering anisotropic elastic tensor from the Neumann-to-Dirichlet map. We show that the shapeof a region where the elastic tensor differs from a known background elastic ten-sor can be detected by a simple monotonicity relation. In addition, we prove aLipschitz stability. Our approach relies on the monotonicity of the Neumann-to-Dirichlet operator with respect to the elastic tensor and the techniques of localizedpotentials.
[1] S. Eberle, B. Harrach, H. Meftahi, and T. Rezgui. Lipschitz stability estimate and recon-struction of Lamé parameters in linear elasticity. arXiv:1906.02194v1, 2019.
86 M5: Inverse Problems for Partial Differential Equations
Minisymposium M6Enumerative and Algebraic Combinatorics
Michael Drmota (Technische Universität Wien)Christian Krattenthaler (Universität Wien)
88 M6: Enumerative and Algebraic Combinatorics
M6a.1
MON15:15
I16:00U407
The Steep-Bounce Zeta Map in Parabolic CatalandWenjie Fang? (Université Paris Est - Marne-la-Vallée)Cesar Ceballos (Universität Wien)Henri Mühle (Technische Universität Dresden)
As a classical object, the Tamari lattice has many generalizations, including ν-Tamari lattices and parabolic Tamari lattices. In this article, we unify these gen-eralizations in a bijective fashion. We first prove that parabolic Tamari lattices areisomorphic to ν-Tamari lattices for bounce paths ν . We then introduce a new com-binatorial object called “left-aligned colorable tree”, and show that it provides a bi-jective bridge between various parabolic Catalan objects and certain nested pairs ofDyck paths. As a consequence, we prove the Steep-Bounce Conjecture using a gen-eralization of the famous zeta map in q, t-Catalan combinatorics. A generalization ofthe zeta map on parking functions, which arises in the theory of diagonal harmonics,is also obtained as a labeled version of our bijection.
[1] N. Bergeron, C. Ceballos, and V. Pilaud, Hopf Dreams, arXiv:1807.03044, 2018.[2] C. Ceballos, W. Fang, and H. Mühle, The Steep-Bounce Zeta Map in Parabolic Cataland,
arXiv:1903.08515, 2019.[3] J. Haglund and N. A. Loehr, A conjectured combinatorial formula for the Hilbert series
fordiagonal harmonics, Discrete Mathematics 298(2005), no. 1-3, 189–204.[4] H. Mühle and N. Williams, Tamari lattices for parabolic quotients of the symmetric
group, arXiv:1804.02761, 2018.
M6a.2
MON16:00
I16:45U407
Some results on characters of symmetric groups
Alexander R. Miller (Universität Wien)
I will present some recent results and conjectures concerning vanishing and divis-
ibility for character values of symmetric groups. The talk will focus on a portion of
underlying combinatorics.
M6: Enumerative and Algebraic Combinatorics 89
M6b.1
MON17:15
I18:00U407
A weight-dependent inversion statistic and Catalan numbersShishuo Fu (Chongqing University)Michael J. Schlosser? (Universität Wien)
We introduce a weight-dependent extension of the inversion statistic, a classi-cal Mahonian statistic on permutations. This immediately gives us a new weight-dependent extension of n!. When we restrict to 312-avoiding permutations, our ex-tension gives rise to a weight-dependent family of Catalan numbers, which happento coincide with the weighted Catalan numbers that were previously introduced byPostnikov and Sagan by weighted enumeration of Dyck paths [4]. While Postnikovand Sagan’s main focus was on the modulo 2 divisibility of the weighted Catalannumbers, we worked out further properties of these numbers that extend those ofthe classical case, such as their recurrence relation, their continued fraction, andHankel determinants. We also discovered an intriguing closed form evaluation ofthe weighted Catalan numbers for a specific choice of weights. We further presentbi-weighted Catalan numbers that generalize Garsia and Haiman’s q, t-Catalan num-bers [2], and again satisfy remarkable properties. These are obtained by refining theweighted Catalan numbers by introducing an additional statistic, namely a weight-dependent extension of Haglund’s bounce statistic [1, 3].
[1] A. Garsia and J.. Haglund, A proof of the q, t-Catalan positivity conjecture, DiscreteMath. 256(3) (2002) 677–717.
[2] A. Garsia and M. Haiman, A remarkable q, t-Catalan sequence and q-Lagrange inver-sion, J. Alg. Combin. 5(3) (1996) 191–244.
[3] J. Haglund, Conjectured statistics for the q, t-Catalan numbers, Adv. Math. 175(2)(2003) 319–334.
[4] A. Postnikov and B.E. Sagan, What power of two divides a weighted Catalan number?,J. Combin. Theory Ser. A 114(5) (2007) 970–977.
90 M6: Enumerative and Algebraic Combinatorics
M6c.1
TUE10:00
I10:45U407
Making Many More Matrix Multiplication MethodsManuel Kauers (Institute for Algebra, Johannes Kepler University)
Strassen’s algorithm for matrix multiplication is based on the observation that theproduct of two 2× 2 matrices can be computed with only 7 multiplications. It isknown that there is no way to do it with fewer multiplications, and that Strassen’smethod is essentially the only way to do it with 7 multiplications. For 3×3-matrices,the situation is less clear. More than 40 years ago, Laderman gave a method thatuses 23 multiplications, and still nobody knows whether there is a way to do it withfewer multiplications. Laderman’s method is not unique. Several authors have laterpresented isolated additional methods that work for arbitrary coefficient rings, andthere is even a family of methods with three free parameters but restricted to coef-ficient rings containing Q. Using SAT solvers and Gröbner bases technology, wehave found more than 10000 pairwise inequivalent new matrix multiplication meth-ods with 23 multiplications. We were able to cluster them together into families withup to 16 parameters which apply without any restriction on the coefficient domain.In conclusion, the set of methods with 23 multiplications appears to be much largerthan it seemed to be until now, and it is quite possible that the methods we found arestill just the tip of the iceberg. Although our results have no immediate implicationson the complexity of matrix multiplication, we do hope that the vast amount of newmethods will eventually contribute to a better understanding of why these methodsexist at all, and whether 23 is best possible for 3×3.
This is joint work with Marijn Heule and Martina Seidl.
M6c.2
TUE10:45
I11:30U407
Local convergence of random planar graphsBenedikt Stufler (Universität Zürich)
The present talk describes the asymptotic local shape of a graph drawn uniformlyat random from all connected simple planar graphs with n labelled vertices. Weestablish a novel uniform infinite planar graph (UIPG) as quenched limit in the localtopology as n tends to infinity. We also establish such limits for random 2-connectedplanar graphs and maps as their number of edges tends to infinity. Our approachencompasses a new probabilistic view on the Tutte decomposition. This allows us tofollow the path along the decomposition of connectivity from planar maps to planargraphs in a uniformed way, basing each step on condensation phenomena for randomwalks under subexponentiality and Gibbs partitions. Using large deviation results forrandom walk, we recover the asymptotic formula by Giménez and Noy [1] for thenumber of planar graphs.
[1] Omer Giménez and Marc Noy. Asymptotic enumeration and limit laws of planar graphs.J. Amer. Math. Soc. 22 (2), 309–329 (2009)
M6: Enumerative and Algebraic Combinatorics 91
M6c.3
TUE11:30
I12:15U407
Exact and asymptotic enumeration of ascent sequences and Fishburn matricesEmma Yu Jin⋆ (Universität Wien)Hsien-Kuei Hwang (Academia Sinica, Taiwan)Michael Schlosser (Universität Wien)
Ascent sequences were introduced by Bousquet-Mélou, Claesson, Dukes, Ki-taev (2010) to unify three seemingly unrelated combinatorial structures: Fishburnmatrices, (2 + 2)-free posets, a family of permutations avoiding a certain pattern,Stoimenow’s involutions, (2-1)-avoiding inversion sequences and non-2-neighbor-nesting matchings.
We present a novel way to recursively decompose ascent sequences, which leadsto a calculation of the Euler-Stirling distribution on ascent sequences, including thenumbers of ascents (asc), repeated entries (rep), zeros (zero) and maximal entries(max). Together with the machinery of basic hypergeometricseries, we are able toprove a bi-symmetry conjecture, which asserts that the quadruples (asc, rep, zero,max) and (rep, asc, max, zero) are equidistributed on ascentsequences.
Furthermore, a direct saddle-point analysis (without relying on any modular forms)is applied to establish the asymptotics of Fishburn numbersand a large number ofothers with a similar sum-of-finite-product form for their (formal) general functions.In addition to solving a conjecture by Jelínek, the application of our saddle-point ap-proach to the distributional aspects is also examined, withmany new limit theoremscharacterized, representing the first of their kind for suchstructures.
[1] G. Andrews and V. Jelínek,On q-series identities related to interval orders, Eur. J.Comb.,39 (2014), 178–187.
[2] M. Bousquet-Mélou, A. Claesson, M. Dukes and S. Kitaev,(2+2)-free posets, ascentsequences and pattern avoiding permutations, J. Combin. Theory Ser. A,117 (2010),884–909.
[3] K. Bringmann, Y.K. Li and R.C. Rhoades,Asymptotics for the number of row-Fishburnmatrices, Eur. J. Combin,41 (2014), 183-196.
[4] P.C. Fishburn,Interval orders and interval graphs: a study of partially ordered sets,John Wiley & Sons, 1985.
[5] V. Jelínek,Counting general and self-dual interval orders, J. Combin. Theory Ser. A,119 (2012), 599–614.
[6] D. Zagier, Vassiliev invariants and a strange identity related to the Dedekind eta-function, Topology, 40 (2001), 945–960.
92 M6: Enumerative and Algebraic Combinatorics
Section S01Logic, Set Theory, and TheoreticalComputer Science
Vera Fischer (Universität Wien)Martin Goldstern (Technische Universität Wien)
94 S01: Logic, Set Theory, and Theoretical Computer Science
S01a.1
MON15:15
I15:45U405
Distributivity of forcing notionsMarlene Koelbing? (Kurt Gödel Research Center, Universität Wien)Vera Fischer (Kurt Gödel Research Center, Universität Wien)Wolfgang Wohofsky (Kurt Gödel Research Center, Universität Wien)
In my talk, I would like to discuss several aspects of the distributivity of forcingnotions, concerning the combinatorial nature and purely forcing-theoretic properties.
S01a.2
MON15:45
I16:15U405
Strong measure zero sets on 2κ
Johannes Philipp Schürz (TU Wien)
The notion of strong measure zero for subsets of the reals R is due to Emil Borel.Call X ⊆ R smz iff
∀(εn)n∈N ∈ RN+ ∃(In)n∈N In ... intervall : ∀n ∈ N λ (In)≤ εn∧X ⊆
⋃n∈N
In.
As R can naturally be identified with 2N, there is a canonical generalization for 2κ ,where κ is a large cardinal, e.g. inaccessible. We call a set X ⊆ 2κ smz iff
∀ f ∈ κκ ∃(ηi)i<κ : ∀i < κ ηi ∈ 2i∧X ⊆
⋃i<κ
[ηi],
where [η ] := x ∈ 2κ : η / x.
I will present basic results about smz sets of R as well as of 2κ , and I want toconclude with a Theorem showing the consistency of
ZFC+ |2κ |= κ++∧ (X ⊆ 2κ is smz ⇔ |X | ≤ κ
+),
relative to the consistency of ZFC+κ is inaccessible.
S01: Logic, Set Theory, and Theoretical Computer Science 95
S01a.3
MON16:15
I16:45U405
The tower spectrumJonathan Schilhan (KGRC, University of Vienna)
Let A,B be infinite sets of natural numbers. Then we say that A is almost containedin B, written as A ⊆∗ B, iff A \B is finite, i.e. all up to finitely many elements of Aare also in B. A tower is a sequence 〈Aα : α < δ 〉 of infinite sets of naturals such thatAβ ⊆∗ Aα for α < β and that is maximal with respect to this property.
Applying Zorn’s lemma we see that towers always exist. But for what regular δ isthere a tower of length δ? The tower spectrum is the set of such δ . We study whatproperties a set C of regular cardinals in a model of GCH has to satisfy to be realizedas the tower spectrum in a forcing extension.
This is joint work with Vera Fischer. The authors would like to thank the AustrianScience Fund, FWF, for generous support through START-Project Y1012-N35.
S01b.1
MON17:15
I18:15U405
Formen des Auswahlaxioms in der RingtheorieLorenz Halbeisen (Department of Mathematics, ETH Zürich)
Zuerst wird gezeigt, dass das Auswahlaxiom äquivalent ist zur Aussage Jeder Ringbesitzt eine Wurzelfunktion. Danach wird untersucht, welche Form des Auswahlax-ioms äquivalent ist zur Aussage Jeder Integritätsbereich hat eine Wurzelfunktionbzw. zur Aussage Jeder Integritätsbereich hat eine n-te Wurzelfunktion. Im letztenTeil wird dann ein Permutationsmodell konstruiert, in dem gewisse dieser Formendes Auswahlaxioms gelten, andere aber nicht.
S01b.2
MON18:15
I18:45U405
Diamonds, games and cardinal invariantsVíctor Torres-Pérez (TU Wien)
On one hand, we prove that WRP and saturation of the ideal NSω1 together imply♦a ∈ [λ ]ω1 : cof(sup(a)) = ω1, for all regular λ ≥ℵ2.
On the other hand, in a joint work with J. Brendle and M. Hrusak, we consider aweak parametrized versions of the diamond principle which game versions of cardi-nal invariants t, u and a. We show that the standard proof that parametrized diamondprinciples prove that the cardinal invariants are small actually show that their gamecounterparts are small. We show that t < tgame and u < ugame are both relativelyconsistent with the ZFC. The corresponding question for a remains open.
96 S01: Logic, Set Theory, and Theoretical Computer Science
S01c.1
TUE10:00
I11:00U405
Automating Resolution is NP-HardMoritz Müller? (Universitat Politècnica de Catalunya, Barcelona)Albert Atserias (Universitat Politècnica de Catalunya, Barcelona)
We show that the problem of finding a Resolution refutation that is at most poly-nomially longer than a shortest one is NP-hard. In the parlance of proof complexity,Resolution is not automatizable unless P = NP. Indeed, we show it is NP-hard todistinguish between formulas that have Resolution refutations of polynomial lengthand those that do not have subexponential length refutations. This also implies thatResolution is not automatizable in subexponential time or quasi-polynomial timeunless NP is included in SUBEXP or QP, respectively.
S01c.2
TUE11:00
I11:30U405
The relation between two weak choice principlesSalome Schumacher (ETH Zürich)
After a brief introduction to permutation models, we will discuss the relationshipbetween the following two weak choice principles:
ACF−:= For every infinite family F of finite sets there is an infinite subfamilyG ⊆F with a choice function.
KWF−:= For every infinite family F of finite sets of size at least two there is aninfinite subfamily G ⊆ F with a Kinna-Wagner selection function. I.e. there is afunction g : G →P (
⋃G ) with /0 6= g(G)( G for all G ∈ G .
In particular it is shown that KWF− does not imply ACF−.
S01: Logic, Set Theory, and Theoretical Computer Science 97
S01c.3
TUE11:30
I12:00U405
Independence and almost disjointnessVera Fischer (University of Vienna)
The cardinal characteristics of the real line arise from various combinatorial prop-erties of the reals. Their study has already a long history and is closely related to thedevelopment of major forcing techniques among which template iterations (see [9]),matrix iterations (see [2], [4]) and creature forcing (see [8]). An excellent introduc-tion to the subject can be found in [1].
Two of classical combinatorial cardinal characteristics of the continuum are thealmost disjointness and independence numbers. An infinite family A of infinitesubsets of N whose elements have pairwise finite intersections and which is maximalunder inclusion, is called a maximal almost disjoint family. The minimal size ofsuch a family is denoted a and is referred to as the almost disjointness number. Afamily A of infinite subsets of N with the property that for any two finite disjointsubfamilies F and G , the set
⋂F\
⋃G is infinite is said to be independent. The
minimal size of a maximal independent family is denoted i and is referred to as theindependence number. The characteristics a and i are among those for which thereare no other known upper bounds apart from c, the cardinality of the continuum. Itis well known that consistently a< i, however both the consistency of i< a and theinequality a≤ i remain open.
In this talk, we will outline some of the major properties of independence andalmost disjointness, describe recent results (see for example [7, 5, 3, 6, 7]) and pointout further interesting open problems.
[1] A. Blass, Combinatorial Cardinal Characteristics of the Continuum, Hand-book of set theory, Springer, Dordrecht, (2010) 395–489.
[2] A. Blass, S. Shelah, Ultrafilters with Small Generating Sets, Israel Journal of Mathe-matics, (1989) 395–465.
[3] J. Brendle, V. Fischer, Y. Khomskii, Definable Independent Families, Proceed-ings of the American Mathematical Society, to appear.
[4] V. Fischer, S. Friedman, D. Mejía, D. Montoya, Coherent Systems of Finite SupportIterations, Journal of Symbolic Logic, 83(1) (2018) 208–236.
[5] V. Fischer, D. Mejía, Splitting, Bounding and Almost Disjointness Can be Quite Differ-ent, Canadian Journal of Mathematics, 69(3) (2017) 502–531.
[6] V. Fischer, D. Montoya, Ideals of Independence, Archive for Mathematical Logic, toappear.
[7] V. Fischer, S. Shelah, The Spectrum of Independence, Archive for Mathematical Logic,to appear.
[8] S. Shelah, On Cardinal Cnvariants of the Continuum, Contemp. Math., 31 (1984), 183–207.
[9] S. Shelah, Two Cardinal Invariants of the Continuum (d < a) and FS Linearly OrderedIterated Forcing, Acta Math., 192(2) (2004), 187–223.
98 S01: Logic, Set Theory, and Theoretical Computer Science
S01c.4
TUE12:00
I12:30U405
Cardinal characteristics and the Borel ConjectureWolfgang Wohofsky? (Kurt Gödel Research Center, Universität Wien)Otmar Spinas (Universität Kiel)
In my talk, I would like to present some tentative considerations in connectionwith a survey article on the Borel Conjecture I recently wrote. I will provide somemodels in which the Borel Conjecture holds, and also models in which the BorelConjecture fails, thereby discussing necessary (but not sufficient) conditions for theBorel Conjecture. I will argue that – given a model for some set-theoretical statementwhich is obtained by a countable support iteration of forcings adding dominatingreals and satisfying the Laver property – it often seems to be easy to modify theconstruction in such a way that the resulting model additionally satisfies the BorelConjecture, but it is not so clear how to make the Borel Conjecture fail in such asituation. As an example, I will mention the model from my joint work with OtmarSpinas in which h= cov(M )< b= s= add(s0) holds (and which is obtained by suchan iteration); in this case it is indeed easy to additionally obtain the Borel Conjecturebut quite unclear how to make it fail.
[1] W. Wohofsky, Borel Conjecture, dual Borel Conjecture, and other variants of the BorelConjecture, in preparation.
[2] O. Spinas, W. Wohofsky, A Sacks amoeba preserving distributivity of P(ω)/ f in, 2019,submitted.
Section S02Algebra, Discrete Mathematics
Clemens Heuberger (Universtität Klagenfurt)Clemens Fuchs (Universtität Salzburg)
100 S02: Algebra, Discrete Mathematics
S02a.1
TUE15:30
I16:00U407
Label-quantities in trees and mappingsAlois Panholzer (Technische Universität Wien)
We consider labelled trees and n-mappings, i.e., functions from the finite set [n] :=1,2, . . . ,n into itself, and study several problems related to the distribution of thelabels along the paths in the tree or in the iteration orbit (i, f (i), f 2(i), . . .) of elementsin a mapping, respectively.
In particular, we take a look at the occurrence and avoidance of certain permutation-patterns in trees and mappings as well as treat problems related to random sequentialadsorption on trees. An analytic combinatorics treatment naturally yields a study ofcertain linear and quasi-linear PDEs.
S02a.2
TUE16:00
I16:30U407
Optimal Multi-Pivot Quicksort and its Asymptotic AnalysisDaniel Krenn? (Alpen-Adria-Universität KlagenfurtΨ)Clemens Heuberger (Alpen-Adria-Universität Klagenfurt)Cecilia Holmgren (Uppsala Universitet)
This talk will be about the asymptotic analysis of the number of comparisonsneeded for sorting a list with multi-pivot quicksort. By the nature of the multi-pivotquicksort, there are many partitioning strategies possible (in contrast to the classicalquicksort, where there is essentially one), and we will learn about the optimal strat-egy. We will focus on the strategy itself, the idea why this is the optimal strategy,and the asymptotic analysis of the number of key comparisons of the correspondingquicksort algorithm (as cost measure for the running time). In the spirit of analyticcombinatorics, the algorithm is "‘translated” to discrete structures and generatingfunctions. The analysis is computer assisted; the main ideas, tools and methods willbe explained.
ΨNote on affiliation (by Sept. 2019): Uppsala Universitet (near past), Universität Salzburg (near future).
S02a.3
TUE16:30
I17:00U407
Galois groups of differential equationsMichael Wibmer? (TU Graz)Annette Bachmayr (Universität Bonn)David Harbater (University of Pennsylvania)Julia Hartmann (University of Pennsylvania)
The Galois theory of differential equations generalizes the classical Galois theoryof polynomials. While the Galois group of a polynomial is a finite group acting onthe roots, the Galois group of a homogeneous linear differential equation is a linearalgebraic group that acts linearly on the vector space of solutions.
The absolute Galois group of the rational function field C(x) is known to be afree profinite group. In this talk I will explain progress towards understanding theabsolute differential Galois group of C(x).
Section S03Number Theory
Timothy Browning (IST Austria)
102 S03: Number Theory
S03a.1
TUE15:30
I16:30U405
Counting points of given degree and bounded height via the height zeta functionKevin Destagnol (IST Austria)
Let X = SymdPn := Pn ×·· ·×Pn/Sd where the symmetric d-group acts by per-muting the d copies of Pn. Manin’s conjecture gives a precise prediction for thenumber of rational points on X of bounded height in terms of geometric invariants ofa resolution of X and the study of Manin’s conjecture for X can be derived from thegeometry of numbers in the cases n > d and for n = d = 2. In this talk, I will explainhow one can use the fact that Pn is an equivariant compactification of an algebraicgroup and the height zeta function machinery in order to study the rational points ofbounded height on X in new cases that are not covered by the geometry of numberstechniques. This might in particular be an interesting testing ground for the latestrefinements of Manin’s conjecture.
S03a.2
TUE16:30
I17:00U405
On Serre’s Problem for ConicsNick Rome? (IST Austria/University of Bristol)Efthymios Sofos (Glasgow University)
We demonstrate an asymptotic formula for the number of ternary diagonal quadraticforms with coefficients of bounded size that have a rational point, thus improvingupon a lower bound due to Hooley [2] and Guo [1], and an upper bound due to Serre[3]. We will discuss the connection between this result and the Loughran–Smeetsconjecture on the number of everywhere locally soluble varieties in a family.
[1] C. R. Guo, On solvability of ternary quadratic forms. Proc. London Math. Soc., 70 ,(1995), 241–263.
[2] C. Hooley, On ternary quadratic forms that represent zero. Glasgow Math. J., 35 ,(1993), 13–23.
[3] J. P. Serre, Spécialisation des éléments de Br2(Q(T1, · · · ,Tn)). C. R. Acad. Sci. ParisSér. I Math., 311 , (1990), 397–402.
S03b.1
TUE17:30
I18:30U405
Arithmetic progressions in binary quadratic forms and norm formsChristopher Frei? (University of Manchester)Christian Elsholtz (TU Graz)
We discuss upper bounds for the length of arithmetic progressions representedby irreducible integral binary quadratic forms (or, more generally, arbitrary normforms), which depend only on the form and the progression’s common difference.
S03: Number Theory 103
S03b.2
TUE18:30
I19:00U405
Solving equations in many variables in primesShuntaro Yamagishi (University of Utrecht, Department of Mathematics)
Let F ∈ Z[x1, . . . ,xn] be a degree d homogeneous form. In 2014, Cook and Mag-yar proved the existence of prime solutions to the equation F(x1, . . . ,xn) = 0 undercertain assumptions on F . In particular, their result requires the number of variablesn to be an exponential tower in d. I will talk about a result related to this work ofCook and Magyar improving on the number of variables required.
104 S03: Number Theory
Section S04Algebraic Geometry, Convex and DiscreteGeometry
Tim Netzer (Universtität Innsbruck)
106 S04: Algebraic Geometry, Convex and Discrete Geometry
S04.1
TUE17:45
I18:15U407
Anchor rings of finite type Gauss map in the Euclidean 3-spaceHassan Al-Zoubi (Al-Zaytoonah University of Jordan)
In 1983 B.-Y. Chen introduced the notion of Euclidean immersions of finite type[1]. A surface S is said to be of finite type corresponding to the first fundamentalform I, or briefly of finite I-type, if the position vector x of S can be written as afinite sum of nonconstant eigenvectors of the Laplacian ∆I , that is,
x = c+k
∑i=1
xi, ∆Ixi = λi xi, i = 1, . . . ,k, (1)
where c is a fixed vector and λ1,λ2, . . . ,λk are eigenvalues of the operator ∆I . Inparticular, if all eigenvalues λ1,λ2, . . . ,λk are mutually distinct, then S is said to beof finite I-type k. When λi = 0 for some i = 1, . . . ,k, then S is said to be of finite nullI-type k. Otherwise, S is said to be of infinite type.
Very little is known about surfaces of finite type in the Euclidean 3-space E3. Ac-tually, so far, the only known surfaces of finite I-type in E3 are the minimal surfaces,the circular cylinders, and the spheres. So in [2] B.-Y. Chen mentions the followingproblem
Problem 1 classify all surfaces of finite Chen I-type in E3.
Many researchers start to solve this problem by investigating important classes ofsurfaces by proving that finite type ruled surfaces, finite type quadrics, finite typecyclides of Dupin and finite type spiral surfaces are surfaces of the only known ex-amples in E3. However, for another classical families of surfaces, such as surfacesof revolution, finite type tubes, translation surfaces as well as helicoidal surfaces,the classification of its finite type surfaces is not known yet.In the present paper, wemainly focus on problem 1 by studying a subclass of tubes in E3, namely anchorrings.
[1] B-Y. Chen, Total mean curvature and submanifolds of finite type, Second edition, WorldScientific Publisher, 2015.
[2] B-Y. Chen, Some open problems and conjectures on submanifolds of finite type, Soo-chow. J. Math., 17 (1991) 161–188.
Section S05Computational Geometry and Topology
Herbert Edelsbrunner (IST Austria)Michael Kerber (Universtität Graz)
108 S05: Computational Geometry and Topology
S05a.1
TUE10:00
I10:30U406
Deep Homological LearningRoland Kwitt (Department of Computer Science, University of Salzburg)
Over the last couple of years, persistent homology, i.e., the prevalent tool in thegrowing field of topological data analysis (TDA), has found many applications in awide variety of scientific areas, ranging from biology, chemistry or medicine to theanalysis of social networks and images.
However, the adoption of concepts from TDA in the machine learning communityis remarkably slow. This is surprising from multiple aspects: First, persistent ho-mology, for instance, offers valuable topological information with potentially highdiscriminative power in supervised learning settings. Second, studying the topol-ogy of intermediate representations learned by today’s neural networks, could offernovel insights into learning progress, “what is actually learned“ and provide a bet-ter understanding of how representation spaces change as input data is sequentiallytransformed through multiple non-linear functions implemented by neural networkcomponents. Third, enforcing certain topological constraints during learning couldhelp machine learning models to generalize better, or avoid highly irregular decisionboundaries.
In the presentation, I will provide an overview of general strategies and prior workon (1) passively leveraging topological features obtained from persistent homology(mostly related to [1, 3]) for supervised learning and (2) present recent advances andchallenges in actively controlling topological characteristics of data in the broadercontext of (unsupervised) learning with neural networks (mostly related to [3, 4]).
[1] C. Hofer, R. Kwitt, M. Niethammer, W. Lin and A. Uhl. Deep Learning with TopologicalSignatures, In NIPS, 2017.
[2] C. Hofer, R. Kwitt, M. Dixit and M. Niethammer. Connectivity Optimized Representa-tion Learning via Persistent Homology, In ICML, 2019.
[3] C. Hofer, R. Kwitt and M. Niethammer. Learning Representations of Persistence Bar-codes. J. Mach. Learn. Res., 2019 (in publication).
[4] C. Hofer, R. Kwitt and M. Niethammer. Graph Filtration Learning. arXiv:1905.10996[cs.LG] (under peer-review), 2019.
S05a.2
TUE10:30
I11:00U406
Topological explorations in neuroscienceDaniela Egas Santander (École polytechnique fédérale de Lausanne)
I will present some of the applications of topology and topological data analysis toneuroscience through an exploration of the collaboration between the applied topol-ogy group at EPFL and the Blue Brain Project. In particular, I will describe how weare using topology to further understand learning or simulations of voltage sensitivedye experiments.
S05: Computational Geometry and Topology 109
S05a.3
TUE11:00
I11:30U406
Spaces of clusteringsAlexander Rolle? (Technische Universität Graz)Luis Scoccola (The University of Western Ontario)
Cluster analysis is the task of finding clusters of points in a dataset; points belong-ing to the same cluster should be similar to one another in some sense, and pointsbelonging to distinct clusters should be less so. In order to accomodate many appli-cations, we use a very general definition: a clustering of a set is just a set of disjointsubsets.
Sometimes a clustering algorithm, rather than producing a single clustering of adataset, produces a set of clusterings. For example, one gets a set of clusterings byrunning a clustering algorithm with a range of parameters. We present methods togive geometric structure to these sets of clusterings, which can be useful for param-eter selection during cluster analysis.
S05a.4
TUE11:30
I12:00U406
Stochastic applications of discrete topologyAnton Nikitenko (IST Austria)
Stochastic geometry studies geometric properties of random objects. Discretetopology provides a toolkit for implying topological properties of discrete pointsets. A reference concept, which unites the two areas, is a subcomplex of the clas-sic Poisson–Delaunay mosaic, the Poisson–Delaunay complex, which encodes thetopology of the union of balls of fixed radius around a Poisson point process. Thetalk is intended to sketch the topological approach in stochastic geometry, and surveysome of the recent developments, in particular deriving the sizes and the distributionof radii of the simplices in the Poisson–Delaunay mosaic using discrete Morse the-ory.
S05a.5
TUE12:00
I12:30U406
Detecting lengths of geodesics with higher-dimensional persistenceŽiga Virk (University of Ljubljana)
Persistent homology is a variant of homology applied to a filtration of a space, thelater often being obtained through the combinatorial construction of Rips or Cechcomplexes at different scales. The usual intuition behind this construction is that itmeasures sizes of holes in the space.
In this talk we will focus on geodesic spaces and describe how the combinatorialproperties of the mentioned complexes allow us to extract relevant geometric in-formation about the underlying space from higher-dimensional persistence. In par-ticular, we can detect certain closed geodesics using higher-dimensional persistenthomology even if geodesics themselves are contractible. We will discuss conditionson local properties of geodesics, which imply this kind of detections.
110 S05: Computational Geometry and Topology
S05b.1
TUE15:30
I16:00U406
Topological Structures of Probabilistic Landscape of Stochastic Reaction Net-worksWei Tian (University of Illinois at Chicago, Chicago, IL 60607, USA)Hubert Wagner (Institute of Science and Technology, Austria, Klosterneuburg 3400, Austria)Anna Terebus (University of Illinois at Chicago, Chicago, IL 60607, USA)Farid Manuchehrfar (University of Illinois at Chicago, Chicago, IL 60607, USA)Herbert Edelsbrunner (Institute of Science and Technology, Austria, Klosterneuburg 3400,Austria)Jie Liang? (University of Illinois at Chicago, Chicago, IL 60607, USA)
Stochasticity plays important roles in fundamental biology such as cellular fatedetermination. Stochasticity arise when small copy number of molecules participatein chemical reactions. The discrete Chemical Master Equation (dCME) provides ageneral framework for investigating stochasticity in reaction systems at mesoscopicscale. With recent advances in exact method (such as ACME) for computing time-evolving and steady state probability landscape by solving the dCMEs [1, 2], a fulltopological treatise [3, 4] of the probability landscapes of stochastic networks is nowfeasible. Such analysis has the promise in yielding new insight into the behavior ofthese stochastic networks. We discuss recent results in topological analysis of proba-bilistic landscape using model systems of Schlögl network, feed-forward loops, andtoggle-switch systems, including computation of persistent homology [5], detectionof topological changes, and topological sensitivity analysis. We also highlight keychallenges in computing persistent homology of high-dimension probability land-scape.
[1] Y. Cao, A. Terebus, and J. Liang. Accurate chemical master equation solution usingmulti-finite buffers. SIAM Multiscale Modeling & Simulation, 14(3):923–963, 2016.
[2] Y. Cao, A. Terebus, and J. Liang. State space truncation with quantified effects foraccurate solution of discrete chemical master equation. Bulletin Math Biology, 78:617–661, 2016.
[3] H. Edelsbrunner, D. Letscher, and A. Zomorodian. Topological persistence and simpli-fication. Discrete and Computational Geometry, 28:511–533, 2002.
[4] H. Edelsbrunner and J. Harer. Computational Topology - An Introduction. AmericanMathematical Society, 2010
[5] H. Wagner. https://bitbucket.org/hubwag/cubicle/.
S05: Computational Geometry and Topology 111
S05b.2
TUE16:00
I16:30U406
On fractal geometric and topologic properties of triangular labyrinth fractalsLigia L. Cristea (Karl Franzens Universität Graz)
Abstract: The results stem from joint work with Paul Surer.Labyrinth fractals were introduced by Cristea and Steinsky [1] as special families
of (self-similar) Sierpinski carpets. Based on the above mentioned concept, we in-troduce triangular labyrinth fractals as the limit of an iterative process: we start withan equilateral triangle of side length one, divide it into m×m equilateral triangles ofside-length 1/m, and then colour some of them in black (indicating the sets that willbe cut out throughout the iterative construction) and the rest in white, according toa given triangular pattern. Then we repeat the process with the remaining white tri-angles. The construction rule is determined by a triangular labyrinth patterns systemthat consists of a pattern (the white pattern) for the “upright” triangles , and one (theyellow pattern) for the “upside down” triangles. We study the topology and frac-tal geometry of triangular labyrinth fractals, with special focus on the arcs (length,dimension, tangents) in the dentrites, and construct, based on the choice of the gen-erating patterns, and the properties induced on their arcs, three families of triangularlabyrinth fractals. Some of these properties are proven by using GIFS as a tool.
[1] L. L. Cristea and B. Steinsky, Curves of infinite length in 4 × 4-labyrinth fractals,Geom. Dedicata, 141:1–17, 2009.
S05b.3
TUE16:30
I17:00U406
Necessary and sufficient conditions for the local unique nearest point propertyof a real sub-manifoldGunther Leobacher? (Universität Graz)Alexander Steinicke (Montanuniversität Leoben)
We investigate the maximal open domain E (M) on which the orthogonal projec-tion map p onto a subset M ⊆ Rd can be defined.
While [Dudek and Holly(1994)] showed that a sufficient condition for the inclu-sion M ⊆ E (M) is that M is a C1 submanifold whose tangent spaces satisfy a localLipschitz condition, we prove that this condition is also necessary. More precisely, ifM is a topological submanifold with M ⊆ E (M), then M must be C1 and its tangentspaces satisfy the same local Lipschitz condition.
The result is of general interest since cleary the above condition is also necessaryfor a topological submanifold to have positive reach, a by now well-known conceptintroduced in [Federer(1959)].
[Dudek and Holly(1994)] E. Dudek and K. Holly. Nonlinear orthogonal projection. AnnalesPolonici Mathematici, 59(1):1–31, 1994.
[Federer(1959)] Herbert Federer. Curvature measures. Trans. Amer. Math. Soc., 93:418–491,1959.
[Leobacher and Steinicke(2019)] R. L. Foote. Existence, Uniqueness and Regularity of theProjection onto Differentiable Manifolds. Preprint, submitted, 2019.
112 S05: Computational Geometry and Topology
S05c.1
TUE17:30
I18:00U406
Convergence and Stability of Random MapperAdam Brown (IST Austria)
In this talk I will discuss a probabilistic theory of convergence between a randommapper graph and the Reeb graph. I will outline the construction of an enhancedmapper graph associated to points randomly sampled from a probability densityfunction concentrated on a constructible R-space. I will then explain how inter-leaving distances for constructible cosheaves and topological estimations via kerneldensity estimates can be used to show that the enhanced mapper graph approximatesthe Reeb graph in a precise way. The content of the talk is joint work with OmerBobrowski, Elizabeth Munch, and Bei Wang.
S05c.2
TUE18:00
I18:30U406
Dualities and clearing for image persistenceUlrich Bauer? (Technical University of Munich (TUM))Maximilian Schmahl (Technical University of Munich (TUM))
Given a filtration of a simplicial complex by simplicial pairs, the inclusions of thepairs induce morphisms in persistent homology and cohomology. We show how toefficiently compute the barcode of the image of this morphism, employing the dualityof persistent homology and cohomology and generalizing the clearing optimizationintroduced by Chen and Kerber. In order to provide the requisite algebraic founda-tions for this generalization, we establish correspondence results between the imagebarcodes for persistent absolute and relative (co)homology. We provide an imple-mentation for Vietoris–Rips filtrations in the framework of the software Ripser.
Section S06Ordinary Differential Equations andDynamical Systems
Peter Szmolyan (Technische Universität Wien)
114 S06: Ordinary Differential Equations and Dynamical Systems
S06.1
WED11:00
I11:30W205
Periodic solutions and and switching behavior in a model of bipolar disorderPeter Szmolyan? (Technische Universität Wien)Ilona Kosiuk (Technische Universität Wien)Ekatarina Kutafina (University Hospital RWTH Aachen)
We analyze a 4-dimensional system of ODEs which models the cyclic behaviorobserved in bipolar disorder. The phenomenological model is based on mutual inhi-bition and cross activation of two states related to mania and depression. The modelappears to be a standard slow-fast system with two slow and two fast componentsdepending on a small parameter ε . However, it turns out that the slow-fast structureis more complicated and several rescalings must be used to capture the global dy-namics. An interesting feature of the problem is that some of the rescaled versionsof the system are non-smooth in the limit ε → 0. This complication is caused bythe occurrence of powers of ε in Michaelis-Menten and Hill terms which model theinteractions in the system. We show that repeated blow-ups are a convenient toolto overcome these difficulties and to prove the existence of a complicated relaxationoscillation.
S06: Ordinary Differential Equations and Dynamical Systems 115
S06.2
WED11:30
I12:00W205
On a system of difference equations of second order solved in a closed fromYacine Halim? (Mila University Center, Algeria)Julius Fergy T. Rabago (University of the Philippines,Philippines )
The term difference equation refers to a specific type of recurrence relation, amathematical relationship expressing xn as some combination of xi with i < n. Theseequations usually appear as discrete mathematical models of many biological andenvironmental phenomena such as population growth and predator-prey interactions,and so these equations are studied because of their rich and complex dynamics. Re-cently, the problem of finding closed-form solutions of rational difference equationsand systems of rational of difference equations have gained considerable interestfrom many mathematicians.
This paper deals with the solution, stability character and asymptotic behavior ofthe rational difference equation
xn+1 =αxn−1 +β
γxnxn−1, n ∈ N0,
where N0 = N∪ 0, α,β ,γ ∈ R+, and the initial conditions x−1 and x0 are nonzero real numbers such that their solutions are associated to generalized Padovannumbers. Also, we investigate the two-dimensional case of the this equation givenby
xn+1 =αxn−1 +β
γynxn−1, yn+1 =
αyn−1 +β
γxnyn−1, n ∈ N0.
[1] Y. Halim and J. F. T. Rabago. On the solutions of a second-order difference equations interms of generalized Padovan sequences. Math. Slovaca., 68: 625–638, 2018.
[2] Y. Halim and M. Bayram. On the solutions of a higher-order difference equation in termsof generalized Fibonacci sequences. Math. Methods Appl. Sci., 39: 2974-2982, 2016.
[3] N. Touafek. On some fractional systems of difference equations. Iranian J Math. Sci.Info., 9: 303–305, 2014.
[4] Y. Yazlik, D. T. Tollu and N. Taskara. On the solutions of difference equation systemswith Padovan numbers. Appl. Math., 12: 15–20, 2013.
116 S06: Ordinary Differential Equations and Dynamical Systems
S06.3
WED12:00
I12:30W205
Optimality conditions of Moreau sweeping processChems Eddine Arroud? (Department of Mathematics, Laboratory LMEPA, University of Jijel,Algeria)Giovanni Colombo (Department of Mathematics, University of Padova, Italy)
We Consider the problem of free endpoint Bolza problem for the controlled Moreauprocess ;Minimize h(x(T )) subject to
x(t) ∈ −NC(t)(x(t))+ f (x(t),u(t))),x(0) = x0 ∈C(0) , u(t) ∈U
(1)
Where C(t) is a closed non-convex moving set, with normal cone NC(t)(x) at x∈C(t),U being the control set and f being smooth.We prove necessary optimality conditions in the form of Pontryagin maximum prin-ciple. We combine techniques from [2] and from [1], a kind of inward/outwardpointing condition is assumed on the reference optimal trajectory at the times wherethe boundary of C(t) is touched.The adjoint equation is not the classical one, a signed vector measure appears andalso the maximality condition is different.
[1] M. Brokate and P. Krejcí, Optimal control of ODE systems involving a rate independentvariational inequality, Discrete and continous dynamical systems serie B. Volume 18(2013), 331-348.
[2] M. Sene, L. Thibault, Regularization of dynamical systems associated with prox-regularmoving sets, Journal of Nonlinear and Convex Analysis 15 (2014), 647-663.
[3] Ch. Arroud and G. Colombo. A Maximum Principle for the controlled Sweeping Pro-cess. Set-Valued Var. Anal (2017), 1-23.
Section S07Partial Differential Equations and Calculusof Variations
Roland Donninger (Universität Wien)
118 S07: Partial Differential Equations and Calculus of Variations
S07a.1
TUE10:00
I11:00W205
Nonlinear asymptotic stability of homothetically shrinking Yang-Mills solitonsBirgit Schörkhuber? (Karlsruhe Institute of Technology)Roland Donninger (University of Vienna)Irfan Glogic (University of Vienna)
We consider the heat flow for Yang-Mills connections on Rd × SO(d) in dimen-sions 5 ≤ d ≤ 9. It is well known that for this model homothetically shrinking soli-tons, i.e. self-similar blowup solutions, exist. In this talk, we discuss the stabilityof the explicitly known Weinkove solution under small SO(d)−equivariant pertur-bations. The case d = 5 was resolved in [1] together with R. Donninger. The gener-alization to higher space dimensions is a recent joint work with I. Glogic.
[1] R. Donninger, B. Schörkhuber Stable blowup for the supercritical Yang-Mills heat flow,arXiv:1604.07737, to appear in Journal of Differential Geometry.
S07a.2
TUE11:00
I11:30W205
Two-locus clines maintained by diffusion and recombination in a heterogeneousenvironmentReinhard Bürger? (Universität Wien)King-Yeung Lam (The Ohio State University, Columbus, OH, USA)Linlin Su (Southern University of Science and Technology, Shenzhen, China)
We study existence and stability of stationary solutions of a system of semilinearparabolic partial differential equations that occurs in population genetics. It describesthe evolution of gamete frequencies in a geographically structured population of mi-grating individuals in a bounded habitat. Fitness of individuals is determined addi-tively by two recombining, diallelic genetic loci that are subject to spatially varyingselection. Migration is modeled by diffusion. Of most interest are spatially non-constant stationary solutions, so-called clines. In a two-locus cline all four gametesare present in the population, i.e., it is an internal stationary solution. We provideconditions for existence and linear stability of a two-locus cline if recombination iseither sufficiently weak or sufficiently strong relative to selection and diffusion. Forstrong recombination, we also prove uniqueness and global asymptotic stability. Forarbitrary recombination, we determine the stability properties of the monomorphicequilibria, which represent fixation of a single gamete. Under specific assumptionson the environment, important characteristics of the cline, such as its slope in thecenter and its asymptotics, can be determined explicitly.
[1] R. Bürger, Two-locus clines on the real line with a step environment,Theor. Popul. Biol. 117 (2017) 1-22
[2] L. Su, K-Y. Lam, R. Bürger, Two-locus clines maintained by diffusion and recombinationin a heterogeneous environment, J. Differential Equations 266 (2019) 7909–7947
S07: Partial Differential Equations and Calculus of Variations 119
S07a.3
TUE11:30
I12:00W205
The negative L2 gradient flow for p-elastic energiesSimon Blatt? (Universität Salzburg)Christopher Hopper (Universität Salzburg)Nicole Vorderobermeier (Universität Salzburg)
This talk will be about ongoing work on the negative L2 gradient flow of p-elasticenergies for curves. After a quick overview over known results for this equation wewill discuss two methods to get short- and longtime existence for these evolutionequations, a regularization method and de Giorgi’s method of minimizing move-ments. We will shortly discuss the pros and cons of both method and present somefuture directions of this research.
S07a.4
TUE12:00
I12:30W205
Small data global regularity for half-wave maps in 4 dimensionsAnna Kiesenhofer? (EPFL)Joachim Krieger (EPFL)
We consider the half-wave maps problem
∂tu = u× (−4)12 u. (1)
where u : Rn+1 −→ S2. It was shown in [1] that for n ≥ 5 this problem is globallywell-posed for smooth initial data which are small in the critical l1 based Besovspace. We extend this result to dimension n = 4.
We can rewrite the half-wave maps equation (1) as a wave maps equation withadditional non-local source terms. To treat these additional nonlinear terms we usean iteration scheme in a suitable space involving the scaling compatible Strichartznorms.
This is a formal analogue of the result by Tataru for wave maps [2].
[1] J. Krieger and Y. Sire Small data global regularity for half-wave maps Anal. PDE 11(2018), no. 3, 661Ã-682.
[2] Tataru, D. Local and global results for wave maps. I. Comm. Partial Differential Equa-tions 23 (1998), no. 9-10, 1781–1793.
120 S07: Partial Differential Equations and Calculus of Variations
S07b.1
TUE15:30
I16:00W205
Threshold for blowup for the supercritical cubic wave equationIrfan Glogic? (University of Vienna)Birgit Schörkhuber (Karlsruhe Institute of Technology)Maciej Maliborski (University of Vienna)
We report on a recent work with B. Schörkhuber and M. Maliborski on the su-percritical focusing cubic wave equation. In [1] we found an explicit, non-trivialself-similar blowup solution u∗T , which is defined on the whole space and exists inall supercritical dimensions d ≥ 5. Furthermore, for d = 7, we analyzed its stabil-ity properties without any symmetry assumptions and proved the existence of a co-dimension one Lipschitz manifold consisting of initial data whose solutions blowupin finite time and converge asymptotically to u∗T . Furthermore, based on numericalsimulations we subsequently performed [2], we conjecture that the stable manifoldof u∗T is in fact a threshold between finite time blowup and dispersion. This find-ing is especially interesting due to its striking similarity to the critical phenomena ingravitational collapse where threshold solutions are typically self-similar.
[1] I. Glogic, B. Schörkhuber, Co-dimension one stable blowup for the supercritical cubicwave equation. preprint, arXiv:1810.07681, October 2018.
[2] I. Glogic, M. Maliborski, and B. Schörkhuber. Threshold for blowup for the supercriticalcubic wave equation. preprint, arXiv:1905.13739, May 2019.
S07b.2
TUE16:00
I16:30W205
On the analyticity of critical points of the Möbius energyNicole Vorderobermeier? (Universität Salzburg)Simon Blatt (Universität Salzburg)
How nice are critical knots of knot energies? We already know that critical knotsof the so-called Möbius energy are smooth. This leads to the question whethercritical knots of the Möbius energy are not only smooth, but also analytic.
In this lecture, we start with giving a short introduction to geometric knot theory.Based on that, we motivate the definition of the Möbius energy and present thetechniques with which the open question on the analyticity was solved. To thebest of our knowledge, this is the first analyticity result in the context of non-localdifferential equations.
[1] S. Blatt and N. Vorderobermeier, On the analyticity of critical points of the Möbiusenergy, Calc. Var. Partial Differential Equations 58 (2019), no. 1, 58:16.
[2] N. Vorderobermeier, On the regularity of critical points for O’Hara’s knot energies:From smoothness to analyticity., Preprint, arxiv:1904.13129, 2019.
S07: Partial Differential Equations and Calculus of Variations 121
S07b.3
TUE16:30
I17:00W205
Unique Continuation for the Zakharov Kuznetsov equationLucrezia Cossetti? (Karlsruhe Institute of Technology, Karlsruhe, Germany)Luca Fanelli (Università di Roma “La Sapienza”, Rome, Italy)Felipe Linares (Instituto Matemática Pura e Aplicada (IMPA), Rio de Janeiro RJ, Brazil)
In this talk we analyze uniqueness properties of solutions to the Zakharov-Kuznetsov(ZK) equation
∂tu+∂3x u+∂x∂
2y u+u∂xu = 0, (x,y) ∈ R2, t ∈ [0,1].
Mainly motivated by the very well known PDE’s counterpart of the Hardy uncer-tainty principle, we provide a two times unique continuation result. More precisely,we prove that given u1,u2 two solutions to ZK, as soon as the difference u1− u2decays (spatially) fast enough at two different instants of time, then u1≡ u2. As ex-pected, it turns out that the decay rate needed to get uniqueness reflects the asymp-totic behavior of the fundamental solution of the associated linear problem. Encour-aged by this fact we also prove optimality of the result.
Some recent results concerning the (3+1)- dimensional ZK equation will be alsopresented.
The seminar is based on a recent paper [1] in collaboration with L. Fanelli and F.Linares.
[1] L. Cossetti, L. Fanelli and F. Linares, Uniqueness results for Zakharov-Kuznetsov equa-tion, Comm. Partial Differential Equations, DOI:10.1080/03605302.2019.1581803
S07c.1
TUE17:30
I18:00W205
Interpolation theory and regularity for incompressible fluid modelsLuigi Forcella? (École Polytechnique Fédérale de Lausanne)Maria Colombo (École Polytechnique Fédérale de Lausanne)Luigi De Rosa (École Polytechnique Fédérale de Lausanne)
Motivated by the precise structure of the operator defining the pressure in an in-compressible fluid model – as the Euler or the Navier-Stokes equations – we willgive some abstract elliptic regularity results for this kind of operator by means of theReal Interpolation Method.
Once established linear and multilinear versions of such theorems, we will showhow these abstract results imply an improvement of regularity in space and in timeof rough solutions to the incompressible models.
S07c.2
TUE18:00
I18:30W205
Missing abstract file directories/Wallauch-DavidSebastian/Wallauch-DavidSebastian-263.pdf !!!
122 S07: Partial Differential Equations and Calculus of Variations
S07c.3
TUE18:30
I19:00W205
Fredholm property of regular hypoelliptic operators in multianisotropic Sobo-lev spacesAni Tumanyan (Russian-Armenian University)
We study conditions under which hypoelliptic operators acting in multianisotropicSobolev spaces in Rn satisfy the Fredholm property. The analysis of the Fredholmproperty of hypoelliptic operators in Sobolev spaces in Rn has certain difficulties -Fredholm theorems for compact manifolds cannot always be used in this case andcharacteristic polynomial of hypoelliptic operator is not homogeneous as in ellipticcase. In this talk we give necessary and sufficient conditions for the Fredholm pro-perty of certain classes of regular hypoelliptic operators with variable coefficientsacting in multianisotropic weighted Sobolev spaces in Rn. Necessary conditions forfulfillment of a priori estimates for differential operators acting in multianisotropicspaces are obtained. The Fredholm property of semielliptic operators, which arethe special subclass of hypoelliptic operators, is studied in the works [1, 2, 3, 4].This work extends the previously obtained results described in the papers [3, 4] fordifferential operators acting in multianisotropic spaces.
[1] G. A. Karapetyan, A. A. Darbinyan, Index of semielliptical operator in Rn. Proceedingsof the NAS Armenia: Mathematics, 42(5) (2007) 33–50.
[2] G. V. Demidenko, Quasielliptic operators and Sobolev type equations. Siberian Mathe-matical Journal, 50(5) (2009) 1060–1069.
[3] A. G. Tumanyan, On the invariance of index of semielliptical operator on the scale ofanisotropic spaces. Journal of Contemporary Mathematical Analysis. 51(4) (2016) 167–178.
[4] A. A. Darbinyan, A. G. Tumanyan, On a priori estimates and Fredholm property of dif-ferential operators in anisotropic spaces. Journal of Contemporary Mathematical Ana-lysis. 53(2) (2018) 61–70.
Section S08Functional Analysis, Real and ComplexAnalysis
Eva Kopecka (Universität Innsbruck)Bernhard Lamel (Universität Wien)
124 S08: Functional Analysis, Real and Complex Analysis
S08a.1
MON15:15
I15:45W206
Infinitesimal symmetries of weakly pseudoconvex manifoldsMartin Kolár (Masaryk University)
We classify the Lie algebras of infinitesimal CR automorphisms of weakly pseu-doconvex hypersurfaces of finite multitype in CN . In particular, we prove that suchmanifolds admit neither nonlinear rigid automorphisms, nor real or nilpotent rota-tions. As a consequence, this leads to a proof of a sharp 2-jet determination resultfor local automorphisms. Moreover, for hypersurfaces which are not balanced, CRautomorphisms are uniquely determined by their 1-jets. The same classification isderived also for special models, given by sums of squares of polynomials. In par-ticular, in the case of homogeneous polynomials the Lie algebra of infinitesimal CRautomorphisms is always three graded. The results also provide a necessary step forsolving the local equivalence problem on weakly pseudoconvex manifolds. This isjoint work with Shin Young Kim (Grenoble).
S08a.2
MON15:45
I16:15W206
The reflection map of sphere mappingsMichael Reiter (University of Vienna)
In this talk we present some recent results about D’Angelo’s reflection map. Weshow how the reflection map can be used to describe biholomorphic invariants ofsphere mappings, such as finite or holomorphic nondegeneracy, and to study in-finitesimal deformations of sphere mappings. In general, for M ⊂ CN and M′ ⊂ CN′
real submanifolds, an infinitesimal deformation of a mapping H : M→M′ is a holo-morphic mapping, defined in a neighborhood of M, whose real part is tangent to M′
along the image of M under H.
S08a.3
MON16:15
I16:45W206
On Fourier multipliersGerhard Racher (Paris Lodron Universität Salzburg)
We show that any locally compact group G containing an open abelian subgroupadmits a nonzero bounded linear map from its von Neumann algebra into L2(G)which intertwines the respective actions by its Fourier algebra. This is a consequenceof the Main Theorem in
T.S. Liu and A.C.M. van Rooij “Translation invariant maps L∞(G) → Lp(G)”.Indag. math. 36 (1974), 306-316.
S08: Functional Analysis, Real and Complex Analysis 125
S08b.1
MON17:15
I17:45W206
On Generic Properties of Nonexpansive MappingsChristian Bargetz (Universität Innsbruck)
We present a few recent results on the generic behaviour of nonexpansive map-pings. The investigated properties include the (local) Lipschitz constant and the con-vergence behaviour of sequences of successive approximations for certain set-valuedmappings.
Connections to the metric geometry of the underlying metric space are also dis-cussed.
S08b.2
MON17:45
I18:15W206
The geometrization of the theory of full Colombeau algebrasEduard Nigsch (Universität Wien)
The theory of Colombeau algebras is an attempt to rigorously define a product ofSchwartz distributions via general regularization methods while retaining as muchcompatibility as possible with distribution theory.
Motivated by applications in low-regularity differential geometry coming fromgeneral relativity, in particular the calculation of the curvature of singular metrics, ageometric (i.e., diffeomorphism invariant) version of the theory has been developedover the last years.
I will give a broad overview of the steps involved in obtaining a diffeomorphisminvariant theory of nonlinear generalized functions and how it is extended to thecase of nonlinear generalized tensor fields. Moreover, I will outline in which waythis structure accomodates virtually all previous variants of Colombeau algebras andhow it can be used to calculate the curvature of conical metrics.
S08c.1
TUE10:00
I10:30W206
Highly irregular separated nets.Michael Dymond (Universität Innsbruck)
In 1998 Burago and Kleiner and (independently) McMullen gave examples ofseparated nets in Euclidean space which are non-bilipschitz equivalent to the integerlattice. We discuss weaker notions of equivalence of separated nets and demonstratethat such notions also give rise to distinct equivalence classes. Put differently, wefind occurrences of particularly strong divergence of separated nets from the integerlattice. Our approach generalises that of Burago and Kleiner and McMullen whichtakes place largely in a continuous setting. Existence of irregular separated nets isverified via the existence of non-realisable density functions ρ : [0,1]d → (0,∞). Inthe presented work, we obtain stronger types of non-realisable densities. This talk isbased on joint work with Vojtech Kaluža.
126 S08: Functional Analysis, Real and Complex Analysis
S08c.2
TUE10:30
I11:00W206
Classification of homogeneous strictly pseudoconvex hypersurfaces in C3
Ilya Kossovskiy (University of Vienna)
Locally homogeneous strictly pseudoconvex hypersurfaces in C2 were classifiedby E. Cartan in 1932. In this joint work with A.Loboda, we complete the classifica-tion of locally homogeneous strictly pseudoconvex hypersurfaces in C3.
S08c.3
TUE11:00
I11:30W206
The ∂ -complex on the Segal-Bargmann spaceFriedrich Haslinger (Universität Wien)
We use the powerful classical methods of the ∂ -complex based on the theory ofunbounded densely defined operators on Hilbert spaces to study certain densely de-fined unbounded operators on the Segal-Bargmann space. These are the annihila-tion and creation operators of quantum mechanics. In several complex variables wehave the ∂ -operator and its adjoint ∂ ∗ acting on (p,0)-forms with coefficients inthe Segal-Bargmann space. We consider the corresponding ∂ -complex and studyspectral properties of the corresponding complex Laplacian = ∂∂ ∗+ ∂ ∗∂ . In ad-dition, we study a more general complex Laplacian D = DD∗ +D∗D, where Dis a differential operator of polynomial type, to find the canonical solutions to theinhomogeneous equations Du = α and D∗v = β .
We also study the ∂ -complex on several models including the complex hyper-bolic space, which turns out to have duality properties similar to the Segal-Bargmannspace (which is common work with Duong Ngoc Son).
S08: Functional Analysis, Real and Complex Analysis 127
S08c.4
TUE11:30
I12:00W206
Sampling, Marcinkiewicz-Zygmund sets, approximation, and quadrature rulesKarlheinz Gröchenig (Universität Wien)
Suppose you are given n samples of a function f on the torus. What can you sayabout f ? How well can you approximate f or the integral
∫ 10 f (x)dx? The key notion
to answer this question is the notion of Marcinkiewicz-Zygmund (MZ) sets [1, 2]. AMZ-set X = (Xn)n∈N for the torus is a sequence of finite sets Xn such that
A‖p‖22 ≤ ∑
x∈Xn
|p(x)|2 ≤ B‖p‖22 ∀p ∈Tn ,
where Tn denotes the subspace of trigonometric polynomials of degree n and A,B >0 are two constants independent of n.
To produce a good approximation of a smooth function f from its samples f |Xn
on an MZ-set, one first constructs a trigonometric polynomial pn ∈ Tn that approx-imates the given samples optimally. Numerically, this amounts to solving a leastsquares problem. For f in the Sobolev space Hσ , one can then prove a convergencerate
‖ f − pn‖= O(n−σ+1/2) .
This also implies an error estimate for quadrature rule that is associated to everyMZ-set.
Using the techniques from [3] one can prove similar results for functions on abounded domain in Rd , where the trigonometric system is replaced by the eigenfunc-tions of the Laplacian, or even more generally on compact Riemannian manifolds.
[1] A. Cohen and G. Migliorati. Optimal weighted least-squares methods. SMAI J. Comput.Math., 3:181–203, 2017.
[2] F. Filbir and H. N. Mhaskar. Marcinkiewicz-Zygmund measures on manifolds. J. Com-plexity, 27(6):568–596, 2011.
[3] K. Gröchenig. Irregular sampling, Toeplitz matrices, and the approximation of entirefunctions of exponential type. Math. Comp., 68(226):749–765, 1999.
128 S08: Functional Analysis, Real and Complex Analysis
S08d.1
TUE15:30
I16:00W206
The heat kernel expansion of Rockland differential operators and applicationsto generic rank two distributions in dimension fiveStefan Haller (University of Vienna)
The short-time heat kernel expansion of elliptic operators provides a link betweenlocal and global features of classical geometries. Many geometric structures re-lated to (non-)involutive distributions can be described in terms of an underlyingfiltered manifold. These include contact structures, Engel manifolds, and all regularparabolic geometries. Filtered analogues of classical (elliptic) operators tend to beRockland, hence hypoelliptic. Remarkably, the heat kernel expansion of Rocklanddifferential operators on general filtered manifolds has the same structure as the onefor elliptic operators. These results have been obtained using a recently introducedcalculus of pseudodifferential operators generalizing the Heisenberg calculus.
In this talk we will illustrate the aforementioned analysis by applying it to anintriguing five dimensional geometry related to the exceptional Lie group G2. Thisgeometry has many names: generic rank two distribution in dimension five, filteredmanifolds of type (2,3,5), or rank two distributions of Cartan type.
[1] Shantanu Dave and Stefan Haller, Graded hypoellipticity of BGG sequences. Preprintavailable at arXiv:1705.01659
[2] Shantanu Dave and Stefan Haller, The heat asymptotics on filtered manifolds. J. Geom.Anal. https://doi.org/10.1007/s12220-018-00137-4 electronically published in January2019 (to appear in print). Preprint available at arXiv:1712.07104
[3] Shantanu Dave and Stefan Haller, On 5-manifolds admitting rank two distributionsof Cartan type. Trans. Amer. Math. Soc. 371(2019), 4911–4929. Preprint available atarXiv:1603.09700
S08: Functional Analysis, Real and Complex Analysis 129
S08d.2
TUE16:00
I16:30W206
A role of Harnack extension in integrating functions over arbitrary subsetsUmi Mahnuna Hanung (Universitas Gadjah Mada, Indonesia)
The Harnack extension discussing a sufficient condition for the integrable func-tions on particular subsets of (a,b) to be integrable on [a,b ] is already included inthe Henstock-Kurzweil integral (see e.g. [1, Theorem 9.22], [2, Corollary 7.11], [3,Theorem 4.4.4]). The Kurzweil-Stieltjes integral reduces to the Henstock-Kurzweilintegral whenever the integrator is an identity function. It is known that if the inte-grator F is discontinuous on [c,d]⊂ [a,b ], then the values of the Kurzweil-Stieltjesintegrals
∫ dc d [F ] g,
∫[c,d] d [F ] g,
∫[c,d) d [F ] g and
∫(c,d) d [F ] g need not coincide (see
[4, Section 5]). This indicates that the Harnack extension principle in the Henstock-Kurzweil integral cannot be valid any longer for the Kurzweil type Stieltjes integra-tion. Moreover, the existence of the integral
∫ ba d[F ]g does not (even in the case of the
identity integrator F(x)≡ x) always imply the existence of the integral∫
T d[F ]g forevery subset T of [a,b ]. This follows from the well known fact that, if e.g. T ⊂ [a,b ]is not measurable, then the existence of the Lebesgue (which is a special case of theHenstock-Kurzweil integral) integral
∫ ba g does not imply, in general, that also the
integral∫
T g exists. Therefore, the aim of this paper is also to resolve the problemof existence of the integral
∫T d[F ]g for every subset T of [a,b ].
[1] R.A. Gordon, The integrals of Lebesgue, Denjoy, Perron, and Henstock, AmericanMathematical Society, USA, 1994.
[2] P.Y. Lee, Lanzhou lectures on Henstock integration, World Scientific Publishing, NewJersey, London & Singapore, 1989.
[3] T.Y. Lee, Henstock-Kurzweil integration on Euclidean spaces, World Scientific Publish-ing, New Jersey, London & Singapore, 2011.
[4] G.A. Monteiro, U.M. Hanung and M. Tvrdý, Bounded convergence theorem for abstractKurzweil-Stieltjes integral, Monatshefte fur Mathematik 180 (2016), 409–434.
130 S08: Functional Analysis, Real and Complex Analysis
S08d.3
TUE16:30
I17:00W206
Intersections with random geodesics in high dimensionsUri Grupel (Universität Innsbruck)
Given a large subset of the sphere A ⊆ Sn−1 does the ratio of lengths between arandom geodesic Γ and the intersection Γ∩A represent the size of A or does it tendto a zero-one law (as the dimension grows)? We will show that for any large set Awe have a distribution that is not concentrated around neither zero nor one.
In contrast to the case of the sphere, for any convex body in high dimensions,we can find a subset, of half the volume, such that the ratio of lengths between theintersection of the random geodesic with the convex body and the intersection of thesubset with the random geodesic will be close to zero-one law.
The analysis of the two cases has different flavors. For the sphere we analyze thesingular values of the Radon transform, in order to bound the variance of the lengthof the random intersection. For convex bodies, we use concentration of measurephenomena.
The results on the sphere can be generalized to the discrete torus or to intersectionon the sphere with higher dimensional subspaces.
Section S09Numerical Analysis, Scientific Computing
Markus Melenk (Technische Universität Wien)Olaf Steinbach (Universität Graz)
132 S09: Numerical Analysis, Scientific Computing
S09a.1
TUE15:30
I16:00
W211/12
Space–Time Finite Element MethodsOlaf Steinbach (TU Graz)
In this talk we will review some recent developments of space–time finite elementmethods for the numerical solution of time–dependent partial differential equationssuch as the heat equation. Here, the variational formulation is considered in thespace–time domain Q, which allows the use of adaptive finite element spaces simul-taneously in space and time, and the use of parallel solution strategies. The firstand more standard approach is based on variational formulations in Bochner spaceswhere unique solvability is based on suitable inf–sup stability conditions. As analternative we may consider the weak solution of the heat equation in anisotropicSobolev spaces which requires the use of appropriate test and ansatz spaces. Herewe introduce a modified Hilbert transformation to end up with a positive definiteand symmetric approximation of the first order time derivative. Several numericalexamples will indicate the potential of the proposed methods.
S09: Numerical Analysis, Scientific Computing 133
S09a.2
TUE16:00
I16:30
W211/12
Vibration of a beam under structural randomnessEberle Robert? (Universität Innsbruck)Michael Oberguggenberger (Universität Innsbruck)
In engineering, the study of vibrations in structures is essential. Some structuressuch as bridges or skis can be modelled as an Euler-Bernoulli beam. The dampedvibrations of an Euler-Bernoulli beam is described by the partial differential equation[1]
(EI(x)wxx)xx +ρA(x)wtt +β (EI(x)(wt)xx)xx +αρA(x)wt = f . (1)
Here, x denotes the longitudinal coordinate, t the time, EI(x) the flexural stiffness,w(x, t) the deflection of the structure in vertical direction, ρ the mass density, A(x)the cross sectional area of the beam, and f the line forces acting on the beam. Thecoefficients β and α stand for the damping coefficients.
The main characterisation of a beam is its bending stiffness EI(x), which can bedetermined by measurements. Such measurements usually yield variations in thebending stiffness EI(x). These variations can be taken into account by adding arandom field r(x) [2] to a mean bending stiffness EI0(x) in the beam equation (1)
EI(x) = EI0(x)+ r(x). (2)
We investigate the vibrations of a ski in the following situation. The ski is clampedin the centre, the ski tip is deflected and released. This experiment accords withthe vibration of a deflected cantilever beam and is modelled by equation (1) withboundary conditions (clamped - free)
w(0, t) = 0, wx(0, t) = 0, wxx(L, t) = 0, wxxx(L, t) = 0 (3)
and initial deflection and velocity:
w(x,0) = w0(x), wt(x,0) = 0. (4)
The initial deflection w0(x) is calculated by the static beam equation. The differ-ential equation is solved by the spectral method [3]. A sequence of simulations isperformed (Monte Carlo simulation) and the vibrations are analysed.
In the future, the presented approach will be applied on other structural problems,such as the vibrations of a composite bridge under traffic load.
[1] R. Eberle, P. Kaps, and M. Oberguggenberger A multibody simulation study of alpineski vibrations caused by random slope roughness, Journal of Sound and Vibration, 446(2019) 225–237.
[2] R. G. Ghanem and P. D. Spanos, Stochastic finite elements: a spectral approach.Springer, New York, 2003.
[3] C. D. Munz and T. Westermann, Numerische Behandlung gewöhnlicher und partiellerDifferenzialgleichungen, Springer, Berlin, 2006.
134 S09: Numerical Analysis, Scientific Computing
S09a.3
TUE16:30
I17:00
W211/12
On some interpolation operators on triangles with curved sidesTeodora Catinas (Babes-Bolyai University, Cluj-Napoca, Romania)
We present some results regarding interpolation operators and linear, positive op-erators defined on triangles having one curved side. They are extensions of the cor-responding operators for functions defined on triangles with straight sides.
The operators defined on domains with curved sides permit essential boundaryconditions to be satisfied exactly and they have important applications in: finite ele-ment methods for differential equations with given boundary conditions, the piece-wise generation of surfaces in CAGD, in obtaining Bezier curves/surfaces in CAGDand in construction of surfaces that satisfy some given conditions.
We consider some Lagrange, Hermite and Birkhoff type interpolation operatorsand Bernstein, Cheney-Sharma and Nielson type operators on triangles with onecurved side. We construct their product and Boolean sum and study their interpola-tion properties.
We study three main aspects of the constructed operators: the interpolation prop-erties, the orders of accuracy and the remainders of the corresponding interpolationformulas.
We use some of the interpolation operators and some of the Bernstein type opera-tors for construction of surfaces that satisfy some given conditions, such as the roofsof the halls.
Finally, we give some numerical examples and we study the approximation errorsfor the operators presented here.
[1] P. Blaga, T. Catinas, G. Coman, Bernstein-type operators on triangle with all curvedsides, Appl. Math. Comput., 218 (2011), 3072–3082.
[2] P. Blaga, T. Catinas, G. Coman, Bernstein-type operators on triangle with one curvedside, Mediterr. J. Math., 9 (2012), No. 4, 843-855.
[3] T. Catinas, P. Blaga, G. Coman, G., Surfaces generation by blending interpolation on atriangle with one curved side, Res. Math., 64 (2013) nos. 3-4, 343-355.
[4] T. Catinas, Extension of some particular interpolation operators to a triangle with onecurved side, Appl. Math. Comput. 315 (2017) 286–297.
[5] G. Coman, T. Catinas, Interpolation operators on a tetrahedron with three curved sides,Calcolo, 47 (2010), no. 2, 113-128.
[6] G. Coman, T. Catinas, Interpolation operators on a triangle with one curved side, BITNumer. Math., 50 (2010), no. 2, 243-267.
S09: Numerical Analysis, Scientific Computing 135
S09b.1
TUE17:30
I18:00
W211/12
an adaptive algorithm for the fractional LaplacianJens Markus Melenk? (TU Wien)Markus Faustmann (TU Wien)Dirk Praetorius (TU Wien)
For the discretization of the integral fractional Laplacian (−∆)s, 0 < s < 1, basedon piecewise linear functions, we present and analyze a reliable weighted residual aposteriori error estimator. In order to compensate for a lack of L2-regularity of theresidual in the regime 3/4 < s < 1, this weighted residual error estimator includesas an additional weight a power of the distance from the mesh skeleton. We proveoptimal convergence rates for an h-adaptive algorithm driven by this error estimatorin the framework of [CFPP14]. Key to the analysis of the adaptive algorithm arenovel local inverse estimates for the fractional Laplacian. These local inverse esti-mates have further applications. For example, they underlie the proof that multileveldiagonal scaling leads to uniformly bounded condition numbers even in the presenceof locally refined meshes.
[CFPP14] C. Carstensen, M. Feischl, M. Page, and D. Praetorius, Axioms of adaptivity, Com-put. Math. Appl. 67 (2014), no. 6, 1195–1253.
S09b.2
TUE18:00
I18:30
W211/12
A new approach of operational matrices for hyperbolic partial differential equa-tionsSomveer Singh? (Department of Mathematics, Indian Institute of Technology Delhi, India)Vineet Kumar Singh (Department of Mathematical Sciences, Indian Institute of Technology,Banaras Hindu University(IIT BHU), Varanasi, India)
A new approach based on operational matrices of Legendre wavelets is introducedfor the class of first order hyperbolic partial differential equations with the giveninitial conditions. Operational matrices of integration of Legendre wavelets are de-rived and utilized to transform the given PDE into the linear system of equations bycombining collocation method. Convergence analysis and error estimation the pre-sented technique are also investigated. Some numerical experiments are performedto demonstrate accuracy and efficiency of the proposed method.
136 S09: Numerical Analysis, Scientific Computing
S09b.3
TUE18:30
I19:00
W211/12
Effect of Impedance on Reflection of Plane Waves in a RotatingMagneto-thermoelastic Solid Half-Space with DiffusionDr. Anand Kumar Yadav ((1) Department of Mathematics, M. M. Deemed To Be UniversityMullana Ambala, India, (2) Shishu Niketan Model Senior Secondary School, Sector 22-D,Chandigarh, India)
In this paper, the governing equations of rotating magneto-thermoelasticity withdiffusion are formulated in context of Lord-Shulman theory of generalized thermo-elasticity. The plane wave solutions of these equations indicate the existence of fourplane waves. The speed of these plane waves are computed for a particular materialand plotted against the diffusion parameter, thermal parameter and frequency. Theplane surface of the half-space is subjected to impedance boundary conditions, wherenormal and tangential tractions are proportional to normal and tangential displace-ment components time frequency, respectively. Reflection of these plane waves froma stress-free thermally insulated surface is studied to obtain the reflection coefficientsand energy ratios of reflected waves. The reflection coefficients and energy ratios ofreflected waves are computed for a particular material and effect of diffusion andangle of incidence is observed graphically on energy ratios. Keywords: ImpedanceBoundary, Thermoelasticity, Diffusion, Plane waves, Reflection Coefficients, Energyratios.
S09c.1
WED11:00
I11:30W206
An operational matrix approach for two-dimensional hyperbolic telegraph equa-tionVinita Devi? (Department of Mathematical Sciences, Indian Institute of Technology, BanarasHindu University(IIT BHU), Varanasi, India)Vineet Kumar Singh (Department of Mathematical Sciences, Indian Institute of Technology,Banaras Hindu University(IIT BHU), Varanasi, India.)
In this work, an efficient method is proposed to find the numerical solution of twodimensional hyperbolic telegraph equation. In this method, the roots of Legendre’spolynomial are taken as the node points for the Lagrange’s polynomial. First, weconvert the main equation in to partial integro-differential equations(PIDEs) with thehelp of initial and boundary conditions. The operational matrices of differentiationand integration are then used to transform the PIDEs in to algebraic generalizedSylvester equations. We compared the results obtained by the proposed method withBernoulli matrix method which shows that the proposed method is accurate for smallnumber of basis function.
S09: Numerical Analysis, Scientific Computing 137
S09c.2
WED11:30
I12:00W206
Multistep finite difference scheme for electromagnetic waves model arising fromdielectric mediaRahul Kumar Maurya? (Department of Mathematical Sciences, Indian Institute of Technol-ogy, Banaras Hindu University(IIT BHU), Varanasi, India)Vineet Kumar Singh (Department of Mathematical Sciences, Indian Institute of Technology,Banaras Hindu University(IIT BHU), Varanasi, India)
In this work, we developed a finite difference scheme for solving two dimen-sional (2D) fractional differential models of electromagnetic waves (FDMEWs) aris-ing from dielectric media. The Caputo’s fractional derivative in time is discretizedby a difference scheme of O(t3−α), 1 < α < 2, and the Laplacian operator is ap-proximated by central difference discretization. Unconditional stability for the pro-posed scheme is established and convergence analysis is done through optimal errorbounds. For 2D FDMEWs, accuracy of O(τ2−α + τ2−β + h2
1 + h22) are proved for
the discrete numerical scheme, where 1 < β < α < 2. Several examples are includedto verify the reliability and computational efficiency of the proposed scheme whichsupport our theoretical findings.
138 S09: Numerical Analysis, Scientific Computing
Section S10Optimization
Radu Bot (Universität Wien)
140 S10: Optimization
S10.1
WED12:00
I12:30W206
A multi-step hybrid conjugate gradient method
Issam A.R. Moghrabi (Gulf University for Science and Technology, Kuwait)
A multi-step hybrid Conjugate Gradient (CG) algorithm is introduced in this research that
generates weighted search directions. The algorithm employs a weighting matrix that is a two-
step quasi-Newton inverse Hessian update matrix that utilizes data from the two most recent
iterations. Multi-step methods have proven to be serious contenders when it comes to the
implementation of quasi-Newton methods, as opposed to the standard class of the quasi-
Newton type that use updates satisfying the standard one-step Secant equation. Our
implementation requires a few additional vectors that need to be updated at each iteration
without requiring storing, coding and updating the Hessian or inverse Hessian matrix
approximations. The numerical performance of the new algorithm is evaluated by comparing it
to other well-known methods developed in a similar context. The new method is tested on a set
of 1941 standard unconstrained optimization problems. Our results reveal encouraging
improvements over other successful methods at a reasonable extra computational as well as
storage costs. The gains are specifically appreciated as the dimension of the problem increases,
thus justifying the extra cost incurred in the implementation of the new method.
Section S11Probability, Statistics
Evelyn Buckwar (Universität Linz)Michaela Szölgyenyi (Universität Klagenfurt)
142 S11: Probability, Statistics
S11a.1
MON15:15
I15:45W205
Computing upper probabilities of failure using Monte Carlo simulation, reweight-
ing techniques and fixed point iteration.
Thomas Fetz (Universität Innsbruck)
In many engineering problems, it can be difficult to specify full probability density
functions for all parameters, due to a lack of data and expert information. In such
cases, we estimate the upper probability p of failure with respect to a parametrized
family of probability density functions ft with t in a parameter set T . The upper
probability p is the solution of the optimization problem p = maxt∈T p(t) where
p(t) is the failure probability for a parameter value t in T .
In our approach, we estimate p(s) for fixed s ∈ T using Monte Carlo simulation.
The algorithm for solving the optimization problem needs values p(t) at many dif-
ferent t which are computed using importance sampling and reweighting techniques
to avoid expensive function evaluations of the performance function (finite element
computations). The accuracy may not be very high if the optimal parameter value t∗
obtained by the optimization procedure is far from the starting value s. In this case,
we iterate which means to simulate again, but now with s = t∗ as fixed parameter
value. The presentation is devoted to such fixed point iterations and in particular to
the reuse and combination of former iteration steps to improve the convergence of
the fixed point iteration, the estimate of the function p and the upper probability p of
failure.
[1] Fetz, T. Efficient computation of upper probabilities of failure. In 12th International
Conference on Structural Safety & Reliability, pages 493–502, 2017.
[2] Fetz T. and M. Oberguggenberger. Imprecise random variables, random sets, and Monte
Carlo simulation. In ISIPTA’15: Proceedings of the Ninth International Symposium on
Imprecise Probability: Theories and Applications, pages 137–146, 2015.
[3] Troffaes, M. C. M. A note on imprecise Monte Carlo over credal sets via importance
sampling. In Proceedings of the Tenth International Symposium on Imprecise Proba-
bility: Theories and Applications, volume 62 of Proceedings of Machine Learning Re-
search, pages 325–332. PMLR, July 2017.
[4] Troffaes, M. C. M., Fetz, T. and Oberguggenberger, M. Iterative importance sampling
for estimating expectation bounds under partial probability specifications. In: Proceed-
ings of the 8th International Workshop on Reliable Engineering Computing. Digital
Repository of the University of Liverpool, pages 139–146, 2018.
S11: Probability, Statistics 143
S11a.2
MON15:45
I16:15W205
BSDEs with Jumps in the Lp-setting: Existence, Uniqueness and ComparisonAlexander Steinicke? (Department of Mathematics and Information Technology, Montanuni-versität Leoben)S. Kremsner (Department of Mathematics and Scientific Computing, KFU Graz)
We study the existence and uniqueness of solutions to backward stochastic dif-ferential equations driven by a Lévy process. In the Lp-setting, p ≥ 2, we find apositive answer, assuming generators that satisfy an extended monotonicity condi-tion and which are not restricted to linear growth in the y-variable. In the delicatecase 1 < p < 2, we show existence and uniqueness of a solution as well, however,the same assumptions on the generator demand new and different proof techniques.In both cases, in dimension 1, if the behavior of the generator’s jump variable meetsan additional, natural condition, then a comparison theorem holds.
144 S11: Probability, Statistics
S11a.3
MON16:15
I16:45W205
Embedding an empirically based decision strategy for backcountry skiers intoa subjective probability framework.Christian Pfeifer? (Universität Innsbruck)Thomas Fetz (Universität Innsbruck)
In the Alps, most fatal snow avalanche accidents are caused by skiers or snow-boarders. As a consequence several decision strategies for backcountry skiers havebeen established in the last two decades in order to prevent avalanche accidents oravalanche fatalities. The most important among them are: Munter’s method of re-duction, Snowcard method (DAV), Stop or Go (OeAV) and an empirically drivendecision strategy [1].
In these strategies the fundamental input variables are
• x = "danger level of avalanche",
• y = "incline of the slope",
• z = "aspect of the slope".
• (skiers behaviour)
To put them into a meaningful mathematical/statistical framework, the strategiesabove could be understood as a "subjective degree of belief" (or subjective risk)causing an avalanche event. Additionally, the skier’s or expert’s perception of dan-ger level (e.g. perception of danger level "considerable/3") , incline/aspect of theslope is more or less imprecise or fuzzy.
In 2009, Pfeifer [2] published an empirical based decision and prediction model,which is based on probabilities of a logistic regression model using variables men-tioned above. We will give a brief intro into this model.
In this application, we use the fuzzy logic framework [3] fuzzifying the empiricalbased method of Pfeifer [2] which is resulting in a decision model within a consistentprobability framework that takes (among others) the subjective human perceptioninto account.
[1] P. Plattner, Werner Munter’s Tafelrunde. Strategische Lawinenkunde auf dem Prüfstand,Berg&Steigen 4/2001.
[2] C. Pfeifer, On probabilities of avalanches triggered by alpine skiers. An empiricallydriven decision strategy for backcountry skiers based on these probabilities, NaturalHazards, 48(3):425-438, 2009.
[3] L. A. Zadeh, Fuzzy sets, Information and Control, 8(3): 338–353, 2009.
S11: Probability, Statistics 145
S11b.1
MON17:15
I17:45W205
Recurrence of 2-dimensional queueing processes,and random walk exit times from the quadrantWolfgang Woess⋆ (TU Graz)Marc Peigné (Université de Tours)
Let X = (X1,X2) be a 2-dimensional random variable and X(n),n ∈ N a sequenceof i.i.d. copies of X . The associated random walk is S(n) = X(1) + X(n). Thecorresponding absorbed-reflected walk W (n),n ∈ N in the first quadrant is givenby W (0) = x ∈ R2
+ and W (n) = max0,W (n − 1)− X(n), where the maximumis taken coordinate-wise. This is often called the Lindley process and models thewaiting times in a two-server queue. We are characterize recurrence of this process,assuming suitable, rather mild moment conditions on X . It turns out that this isdirectly related with the tail asymptotics of the exit time of the random walk x+S(n)from the quadrant, so that the main part of this paper is devoted to an analysis of thatexit time in relation with the drift vector, i.e., the expectation of X .
S11b.2
MON17:45
I18:15W205
Random walk on the random connection modelErcan Sönmez (University of Klagenfurt )
We study the behavior of the random walk in a continuum independent long-rangepercolation model, in which two given vertices x and y are connected with probabilitythat asymptotically behaves like |x−y|−α with α > d, where d denotes the dimensionof the underlying Euclidean space. Focus is on the random connection model inwhich the vertex set is given by the realization of a homogeneous Poisson pointprocess. We show that this random graph exhibits the same properties as classicaldiscrete long-range percolation models with regard to recurrence and transience ofthe random walk. We fully characterize recurrence and give sufficient conditions fortransience.
146 S11: Probability, Statistics
Section S12Financial and Actuarial Mathematics
Gunther Leobacher (Universität Graz),Stefan Thonhauser (Universität Graz)
148 S12: Financial and Actuarial Mathematics
M2c.3
THU18:30
I19:00W205
Interacting Particle Systems, Default Cascades and the M1-topologyStefan Rigger? (Universität Wien)Christa Cuchiero (Wirtschaftsuniversität Wien)Sara Svaluto-Ferro (Universität Wien)
The M1-topology is one of the lesser-known topologies originally introduced bySkorokhod. An advantageous feature of this topology (in comparison to the morewidely used J1-topology) is that it is particularly well-suited to deal with monotonefunctions. We prove a tightness result for processes that can be decomposed into acontinuous and a monotone part and discuss its relevance for particle systems withsingular interactions, which serve as models for networks with contagious effects.Applications include a variety of phenomena, such as neuronal synchronization inthe brain, systemic risk of financial lending networks, and the supercooled Stefanproblem.
S12: Financial and Actuarial Mathematics 149
S12.1
TUE10:00
I10:30U210
Utility Maximizing Strategies under Increasing Model Uncertainty in a Multi-variate Black Scholes Type MarketJörn Sass? (University of Kaiserslautern)Dorothee Westphal (University of Kaiserslautern)
When modeling financial markets one is often confronted with model uncertaintyin the sense that parameters of the model or the distributions of some factors in themodel are only known up to a certain degree. We look at a multivariate continuous-time financial market driven by a Brownian motion.
In an extreme setting, we investigate how optimal trading strategies for a util-ity maximization problem behave when we have Knightian uncertainty on the drift,meaning that the only information is that the drift parameter lies in a certain set.This is a robust portfolio optimization setting in which we aim at the best perfor-mance given that the true drift parameter is the worst possible parameter for ourchosen strategy within this set. If the model uncertainty exceeds a certain threshold,simple strategies such as uniform portfolio diversification outperform more sophis-ticated ones due to being more robust, see [1] in discrete time. We extend theseresults to continuous time, provide quite explicit strategies and convergence results,and discuss these in terms of different performance measures.
Another approach to deal with uncertainty is the use of expert opinions in a para-metric model. In such a market model with with Gaussian drift, the underlying driftcan be estimated by the Kalman filter from the observed stock returns and expertopinions can lead to an improvement of the basic filter. So the expert opinions re-duce parameter uncertainty in this setting. In [2] bounds on the performance arederived. We relate and compare the two approaches.
[1] G. Ch. Pflug, A. Pichler, D. Wozabel, The 1/N investments strategy is optimal under highmodel ambiguity, Journal of Banking & Finance 36 (2012) 410–417.
[2] J. Sass, D. Westphal, R. Wunderlich, Expert opinions and logarithmic utility maximiza-tion for multivariate stock returns with Gaussian drift, International Journal of Theoret-ical and Apllied Finance 20 (2017) 41 pages.
150 S12: Financial and Actuarial Mathematics
S12.2
TUE10:30
I11:00U210
Severity Modeling of Extreme Insurance Claims for TarifficationSascha Desmettre? (Karl-Franzens-Universität Graz)Christian Laudagé (Fraunhofer Institut für Techno- und Wirtschaftsmathematik)Jörg Wenzel (Fraunhofer Institut für Techno- und Wirtschaftsmathematik)
Generalized linear models (GLMs) are common instruments for the pricing ofnon-life insurance contracts. Among other things, they are used to estimate the ex-pected severity of insurance claims. However, these models do not work adequatelyfor extreme claim sizes. To accomodate for these, we develop the threshold sever-ity model in [1], that splits the claim size distribution in areas below and above agiven threshold. More specifically, the extreme insurance claims above the thresh-old are modeled in the sense of the peaks-over-threshold (POT) methodology fromextreme value theory using the generalized Pareto distribution for the excess dis-tribution, and the claims below the threshold are captured by a generalized linearmodel based on the truncated gamma distribution. The threshold severity modelfor the first time combines the POT modeling for extreme claim sizes with GLMsbased on the truncated gamma distribution. Based on this framework, we developthe corresponding log-likelihood function w.r.t. right-censored claim sizes, which isa typical issue that arises for instance in private or car liability insurance contracts.Finally, we demonstrate the behavior of the threshold severity model compared tothe commonly used generalized linear model based on the gamma distribution in thepresence of simulated extreme claim sizes following a log-normal as well as BurrType XII distribution.
[1] C. Laudagé, S. Desmettre & J. Wenzel, Severity modeling of extreme insurance claimsfor tariffication, Insurance: Mathematics and Economics, Volume 88, Pages 77-92,https://doi.org/10.1016/j.insmatheco.2019.06.002, (2019).
S12: Financial and Actuarial Mathematics 151
S12.3
TUE11:00
I11:30U210
Optimal dividends and funding in risk theoryJosef Anton Strini? (Graz University of Technology)Stefan Thonhauser (Graz University of Technology)
We extend the classical dividend maximization problem from ruin theory by intro-ducing a funding opportunity in addition to the existing dividend control. This newcontrol is, in contrast to the dividend payments, able to increase the reserve process.Our goal is to reveal a suitable control strategy which maximizes the difference ofexpected cumulated discounted dividends and total expected discounted additionalfunding (subject to some proportional transaction costs). As presented in [1], div-idends are modelled using the common approach, whereas for the funding oppor-tunity we use the jump times of another independent Poisson process at which wechoose an appropriate funding height, contrary to classical capital injection strate-gies, see [2]. In case of exponentially distributed claims we are able to determine anexplicit solution to the problem and derive an optimal strategy whose nature heavilydepends on the size of the transaction costs. On top of this, the optimal strategyidentifies unfavourable surplus positions prior to ruin at which refunding is highlyrecommended. Some numerical illustrations conclude our contribution, where wedepict the optimal strategy as a function of the transaction cost parameter.
[1] P. Azcue, N. Muler, Stochastic optimization in insurance, Springer Briefs in QuantitativeFinance, Springer, New York, 2014.
[2] N. Kulenko, H. Schmidli, Optimal dividend strategies in a Cramér-Lundberg model withcapital injections, Insurance Math. Econom., 43(2), (2008) 270–278.
S12.4
TUE11:30
I12:00U210
Optimal reinsurance for expected penalty functions in risk theoryStefan Thonhauser? (Technische Universität Graz)Michael Preischl (Technische Universität Graz)
Complementing existing results on minimal ruin probabilities, we minimize ex-pected discounted penalty functions (or Gerber-Shiu functions) in a Cramér-Lundbergmodel by choosing optimal reinsurance. Reinsurance strategies are modeled as timedependent control functions, which lead to a setting from the theory of optimalstochastic control and ultimately to the problem’s Hamilton-Jacobi-Bellman equa-tion. We show existence and uniqueness of the solution found by this method andprovide numerical examples involving light and heavy tailed claims and also give aremark on the asymptotics.
152 S12: Financial and Actuarial Mathematics
List of Participants
154 List of Participants
Agaltsov Alexey (Göttingen)Al-Zoubi Hassan Mahmoud (Amman)Arroud Chems eddine (Jijel)Auzinger Winfried (Wien)
Bag Ömer Faruk (Wien)Bargetz Christian (Innsbruck)Bauer Ulrich (Garching)Behrndt Jussi (Graz)Blatt Simon (Salzburg)Brown Adam (Klosterneuburg)Browning Timothy Daniel
(Klosterneuburg)Buckwar Evelyn (Linz)Bura Efstathia (Wien)Bátkai András (Feldkirch)Bürger Reinhard (Wien)
Catinas Teodora (Cluj-Napoca)Corbet René (Graz)Cossetti Lucrezia (Karlsruhe)Cristea Ligia Loretta (Graz)Cuchiero Christa (Wien)
Desmettre Sascha (Graz)Destagnol Kevin (Klosterneuburg)Devi Vinita
(Lansdowne, Uttarakhand)Donninger Roland (Vienna)Dorner Christian (Hirtenberg)Drmota Michael (Wien)Dymond Michael (Innsbruck)
Eberle Sarah (Frankfurt am Main)Eberle Robert (Innsbruck)Edelsbrunner Herbert
(Klosterneuburg)Egas Santander Daniela (LAUSANNE)Eigenthaler Günther (Wien)Erath Christoph (Darmstadt)
Fang Wenjie (Marne-la-Vallée)Feilhauer Thomas (Dornbirn)Fetz Thomas (Innsbruck)Finck Steffen (Dornbirn)Fischer Vera (Vienna)Forcella Luigi (Lausanne)Frei Christopher (Birchwood)Frick Klaus (Buchs)Fugacci Ulderico (Torino)
Gesztesy Fritz (Waco)Glogic Irfan (Vienna)Goldstern Martin (Wien)Gonon Lukas (St. Gallen)Groechenig Karlheinz (Wien)Grupel Uri (Innssbruck)Gröblinger Ortrun (Innsbruck)Gschwandtner Philipp (Innsbruck)
Halbeisen Lorenz (Zuerich)Halim Yacine (Mila)Haller Stefan (Vienna)Haltmeier Markus (Innsbruck)Hanung Umi Mahnuna (Yogyakarta)Haslinger Friedrich (Wien)Hellwig Michael (Dornbirn)Hettlich Frank (Karlsruhe)Heuberger Clemens (Klagenfurt)Hochbruck Marlis (Karlsruhe)Hofbauer Josef (Wien)Hofer Martin (Wien)Huber Stefan (Puch/Salzburg)
Imamoglu Ozlem (Zurich)
Jawecki Tobias (Wien)Jin Emma Yu (Vienna)
Kadunz Gert (Klagenfurt)Kaltenbacher Barbara (Klagenfurt)Kappeler Thomas (Zurich)
List of Participants 155
Kauers Manuel (Linz)Kepplinger Peter (Dornbirn)Kerber Michael (Graz)Khosrawi Wahid (Zürich)Kiesenhofer Anna (Lausanne)Klein Rebecca (Saarbrücken)Koelbing Marlene (Wien)Kolar Martin (Brno)Kopecka Eva (Innsbruck)Kossovskiy Ilya (Vienna)Kostenko Aleksey (Vienna)Kowar Richard (Innsbruck)Kratsios Anastasis (Zürich)Krattenthaler Christian (Wien)Krenn Daniel (Klagenfurt)Kwitt Roland (Salzburg)
Lamel Bernhard (Wien)Lenz Daniel (Jena)Leobacher Gunther (Graz)Liang Jie (Chicago)Liebrecht Michael (Linz)Lindner Andreas
(Bad Ischl, Österreich)Lischka Marc (Zürich)
Maier Andreas (Erlangen)Maurya Rahul Kumar (Varanasi)Melenk Markus (Wien)Mercier Gwenael (Wien)Michor Johanna (Wien)Miller Alexander R. (Wien)Moghrabi Issam A.R. (KUWAIT)Müller Oskar (Dornbirn)Müller Moritz (Barcelona)
Netzer Tim (Innsbruck)Neumann Lukas (Innsbruck)Nigmetov Arnur (Graz)Nigsch Eduard (Vienna)Nikitenko Anton (Klosterneuburg)
Oberguggenberger Michael(Innsbruck)
Obmann Daniel (Innsbruck)
Panholzer Alois (Wien)Pfeifer Christian (Innsbruck)Pickhardt Carola (Sigmaringen)Piorkowski Mateusz (Vienna)Piotrowska-Kurczewski Iwona
(Bremen)Plankensteiner Kathrin (Dornbirn)Pottmann Helmut (Thuwal)
Racher Gerhard (Salzburg)Ramlau Ronny (Linz)Rao Ziping (Vienna)Raykov Gueorgui Dimitrov
(Santiago de Chile)Razen Gisela Maria (Feldkirch)Reiter Michael (Vienna)Rheinberger Klaus (Dornbirn)Rigger Stefan (Vienna)Rolle Alexander Scott (Graz)Rome Nick (Klosterneuburg)Romero Hinrichsen Francisco (Graz)Rothermel Dimitri (Saarbrücken)Rundell William
(College Station, Texas)
SINGH SOMVEER (NEW DELHI)Sakhnovich Alexander (Wien)Sass Jörn (Kaiserslautern)Schelling Andreas (Spittal)Schilhan Jonathan (Vienna)Schlosser Michael Joseph (Wien)Schreiber Hannah (Graz)Schumacher Salome (Zürich)Schwab Johannes (INNSBRUCK)Schörkhuber Birgit (Karlsruhe)Schürz Johannes Philipp (Wien)Sellars Philip (Cambridge)
156 List of Participants
Shoukourian Hayk(Garching bei München)
Sigmund Karl (Vienna)Sousa Rúben (Porto)Steinbach Olaf (Graz)Steinicke Alexander (Leoben)Stockinger Harald (Wien)Strini Josef Anton (Graz)Stufler Benedikt (Zurich)Stutterecker Werner (Pinkafeld)Svaluto-Ferro Sara (Wien)Szmolyan Peter (Wien)Szölgyenyi Michaela (Klagenfurt)Sönmez Ercan (Klagenfurt)Süss-Stepancik Evelyn (Baden)
Teichmann Josef (Zurich)Terracini Susanna (Torino)Teschl Gerald (Wien)Teschl Susanne (Vienna)Thaller Bernd (Graz)Thonhauser Stefan (Graz)Tichy Robert (Graz)Torres-Perez Victor (Wien)Tumanyan Ani (Yerevan)
Unterkofler Karl (Dornbirn)
Virk Ziga (Ljubljana)Vorderobermeier Nicole (Salzburg)
Wald Anne (Saarbrücken)Wallauch David Sebastian (Vienna)Weber Frank (Dornbirn)Wibmer Michael (Graz)Winkler Reinhard (Vienna)Woess Wolfgang (Graz)Wohofsky Wolfgang (Wien)Wutte Hanna Sophia (Zurich)
Yadav Anand Kumar (Panchkula)
Yamagishi Shuntaro (Utrecht)
Zangerl Gerhard (Innsbruck)Zastrow Andreas (Gdansk)Zotlöterer Franz (St. Pölten)
schlag wilhelm (New Haven)
Index
Agaltsov, Alexey, 80, 154Aichinger, Michael, 54Al-Zoubi, Hassan Mahmoud, 106,
154Alberti, Giovanni, 84Ammari, Habib, 84Antholzer, Stephan, 55, 78Arroud, Chems Eddine, 116Arroud, Chems eddine, 154Atserias, Albert, 96Auzinger, Winfried, 154Aviles-Rivero, Angelica, 55
Bürger, Reinhard, 21Bátkai, András, 21Bachmayr, Annette, 100Bag, Ömer Faruk, 154Balemi, Silvano, 51Bargetz, Christian, 125, 154Batkái, András, 23Bauer, Ulrich, 112, 154Baum, Ann-Kristin, 54Behrndt, Jussi, 23, 71, 154Beiglboeck, Mathias, 21Blatt, Simon, 119, 120, 154Bluder, Olivia, 53Bot, Radu, 22, 139Brown, Adam, 112, 154Browning, Timothy, 22, 101Browning, Timothy Daniel, 154
Bruneau, Vincent, 74Buckwar, Evelyn, 22, 141, 154Bürger, Reinhard, 118, 154Bura, Efstathia, 9, 43, 44, 154Bátkai, András, 154
Catinas, Teodora, 134, 154Ceballos, Cesar, 88Colombo, Maria, 121Corbet, René, 154Cossetti, Lucrezia, 121, 154Cristea, Ligia Loretta, 111, 154Cuchiero, Christa, 59, 61–63, 148,
154
Dür, Gabriela, 29De Rosa, Luigi, 121Desmettre, Sascha, 150, 154Destagnol, Kevin, 102, 154Devi, Vinita, 136, 154Donninger, Roland, 22, 73, 76, 117,
118, 154Dorner, Christian, 154Drmota, Michael, 23, 87, 154Dymond, Michael, 125, 154
Eberle, Robert, 133, 154Eberle, Sarah, 85, 154Edelsbrunner, Herbert, 22, 107, 110,
154Egas Santander, Daniela, 108, 154
158 Index
Egorova, Iryna, 74, 75Eichmair, Michael, 21Eigenthaler, Günther, 154Eisenle, Tanja, 29Erath, Christoph, 50, 154
Fanelli, Luca, 121Fang, Wenjie, 88, 154Faustmann, Markus, 135Feilhauer, Thomas, 68, 154Fetz, Thomas, 10, 21, 142, 154Finck, Steffen, 21, 22, 65, 68, 154Fischer, Vera, 22, 93, 94, 97, 154Forcella, Luigi, 121, 154Frei, Christopher, 9, 43, 44, 102, 154Frick, Klaus, 51, 154Fu, Shishuo, 89Fuchs, Clemens, 10, 22, 99Fugacci, Ulderico, 154
Gérard, Patrick, 72Gesztesy, Fritz, 9, 43, 44, 154Glogic, Irfan, 118, 120Glogic, Irfan, 73, 154Goldstern, Martin, 22, 93, 154Gonon, Lukas, 58, 59, 154Grigoryeva, Lyudmila, 58, 59Gröblinger, Ortrun, 154Groechenig, Karlheinz, 127, 154Grupel, Uri, 154Gschwandtner, Philipp, 67, 154Guerra, Manuel, 75
Halbeisen, Lorenz, 95, 154Halim, Yacine, 115, 154Haller, Stefan, 128, 154Haltmeier, Markus, 22, 49, 51–53, 55,
78, 154Hanung, Umi Mahnuna, 129, 154Harbater, David, 100Harrach, Bastian, 85
Hartmann, Julia, 100Haslinger, Friedrich, 126, 154Heiss, Jakob, 60Hellwig, Michael, 21, 154Hettlich, Frank, 80, 154Heuberger, Clemens, 10, 22, 99, 154Hochbruck, Marlis, 9, 43, 45, 154Hofbauer, Josef, 9, 43, 45, 154Hofer, Martin, 154Hopper, Christopher, 119Huber, Stefan, 154
Imamoglu, Ozlem, 46, 154Imamoglu, Özlem, 9, 43
Jawecki, Tobias, 154Jin, Emma Yu, 91, 154
Kadunz, Gert, 154Kaltenbacher, Barbara, 13, 21, 23, 53,
77, 79, 81, 154Kappeler, Thomas, 72, 154Kauers, Manuel, 90, 155Kepplinger, Peter, 69, 155Kerber, Michael, 22, 107, 155Khosrawi, Wahid, 62, 155Kiesenhofer, Anna, 119, 155Klein, Rebecca, 84, 155Koelbing, Marlene, 94, 155Kolar, Martin, 124, 155Kopecka, Eva, 22, 123, 155Kossovskiy, Ilya, 126, 155Kostenko, Aleksey, 72, 155Kowar, Richard, 51, 155Krabichler, Thomas, 63Kratsios, Anastasis, 60, 155Krattenthaler, Christian, 23, 87, 155Kremsner, Stefan, 143Krenn, Daniel, 100, 155Krieger, Joachim, 119Kukharenko, Oleksandra, 59
Index 159
Kwitt, Roland, 108, 155
Lam, King-Yeung, 118Lamel, Bernhard, 22, 123, 155Laudagé, Christian, 150Lenz, Daniel, 72, 155Leobacher, Gunther, 22, 111, 147,
155Li, Housen, 78Liang, Jie, 110, 155Liebrecht, Michael, 54, 155Linares, Felipe, 121Lindner, Andreas, 155Lischka, Marc, 155
MaaSS, Peter, 83Maier, Andreas, 56, 155Maliborski, Maciej, 120Manuchhrfar, Farid, 110Maurya, Rahul Kumar, 137, 155Meftahi, Houcine, 85Melenk, Markus, 22, 131, 135, 155Mercier, Gwenael, 82, 155Michor, Johanna, 74, 155Miller, Alexander R., 88, 155Moghrabi, Issam A.R., 140, 155Mühle, Henri, 88Müller, Moritz, 96, 155Müller, Oskar, 155
Netzer, Tim, 22, 105, 155Neumann, Lukas, 53, 155Nguyen, Linh, 51Nguyen, Linh V., 52Nguyen, Tram Thi Ngoc, 81Nigmetov, Arnur, 155Nigsch, Eduard, 125, 155Nikitenko, Anton, 109, 155Nuster, Robert, 52
Oberguggenberger, Michael, 133,155
Obmann, Daniel, 78, 155Ortega, Juan-Pablo, 58, 59Ostermann, Alexander, 21
Panholzer, Alois, 100, 155Parteder, Erik, 54Pedretscher, Barbara, 53Peigne’, Marc, 145Pfeifer, Christian, 155Pickhardt, Carola, 155Piorkowski, Mateusz, 75, 155Piotrowska-Kurczewski, Iwona, 83,
155Plankensteiner, Kathi, 21Plankensteiner, Kathrin, 155Platonina, Carmen, 21Pottmann, Helmut, 9, 13, 43, 46, 155Praetorius, Dirk, 135Preischl, Michael, 151PreiSSinger, Markus, 69
Rabanser, Simon, 53Racher, Gerhard, 124, 155Raikov, Georgi, 74Ramlau, Ronny, 78, 155Rao, Ziping, 76, 155Raykov, Gueorgui Dimitrov, 155Razen, Gisela Maria, 155Reiter, Michael, 124, 155Rezgui, Taher, 85Rheinberger, Klaus, 69, 155Rigger, Stefan, 148, 155Rolle, Alexander Scott, 109, 155Rome, Nick, 102, 155Romero Hinrichsen, Francisco, 84,
155Rosskopf, Ramona, 69Rothermel, Dimitri, 54, 155Rundell, William, 23, 77, 79, 155
Sakhnovich, Alexander, 73, 155
160 Index
Sass, Jörn, 149, 155Schelling, Andreas, 155Schilhan, Jonathan, 95, 155Schlag, Wilhelm, 9, 13, 43, 47schlag, wilhelm, 156Schlosser, Michael Joseph, 89, 155Schmahl, Maximilian, 112Schoenlieb, Carola, 55Schörkhuber, Birgit, 118, 120, 155Schröcker, Hans-Peter, 21Schreiber, Hannah, 155Schürz, Johannes Philipp, 94, 155Schumacher, Salome, 96, 155Schuster, Thomas, 54, 81, 84Schwab, Johannes, 55, 78, 155Sellars, Philip, 55, 155Senoner, Natascha, 21Sfakianaki, Georgia, 83Shoukourian, Hayk, 66, 156Sigmund, Karl, 9, 156SINGH, SOMVEER, 155Singh, Somveer, 135Sönmez, Ercan, 145, 156Sousa, Rúben, 75, 156Spinas, Otmar, 98Steinbach, Olaf, 22, 131, 132, 156Steinicke, Alexander, 111, 143, 156Stockinger, Harald, 156Strini, Josef Anton, 151, 156Stufler, Benedikt, 90, 156Stutterecker, Werner, 156Su, Linlin, 118Süss-Stepancik, Evelyn, 156Svaluto-Ferro, Sara, 63, 148, 156Szmolyan, Peter, 22, 113, 114, 156Szölgyenyi, Michaela, 156Szölgyenyi, Michaela, 22, 141
Teichmann, Josef, 22, 57, 59–62, 156Terebus, Anna, 110Terracini, Susanna, 9, 43, 48, 156Teschl, Gerald, 21, 23, 71, 74, 75, 156Teschl, Susanne, 23, 156Thaller, Bernd, 23, 156Thonhauser, Stefan, 22, 147, 151, 156Tian, Wei, 110Tichy, Robert, 21, 156Torres-Perez, Victor, 95, 156Tumanyan, Ani, 122, 156
Unterkofler, Karl, 13, 21, 23, 156
Virk, Ziga, 109, 156Vorderobermeier, Nicole, 119, 120,
156
Wagner, Hubert, 110Wahl, Roman, 54Wald, Anne, 81, 84, 156Wallauch, David Sebastian, 156Weber, Frank, 156Wenzel, Jörg, 150Westphal, Dorothee, 149Wibmer, Michael, 100, 156Winkler, Reinhard, 156Wintz, Timothée, 84Woess, Wolfgang, 145, 156Wohofsky, Wolfgang, 94, 98, 156Wutte, Hanna Sophia, 60, 156
Yadav, Anand Kumar, 136, 156Yakubovich, Semyon, 75Yamagishi, Shuntaro, 103, 156
Zangerl, Gerhard, 52, 156Zastrow, Andreas, 156Zotlöterer, Franz, 156